Article 7102 of comp.org.decus: The DECUServe Journal --------------------- December, 1997 From the Editors' Keyboard . . . . . . . . . . . . . 2 What's inside; back to the basics; a gentle reminder DCL PIPE Magic . . . . . . . . . . . . . . . . . . . 3 Defining a symbol inside a pipe expression About AMDS . . . . . . . . . . . . . . . . . . . . . 8 A tool you might already have and not know about When Faster Networks Don't Help . . . . . . . . . 11 Some field experiences with high-speed LANs DECWindows Won't Start for User . . . . . . . . . 16 Why DECWindows starts for SYSTEM but not for a user MicroVAX Terminal Ports . . . . . . . . . . . . . 19 Problems to check for, and how to dodge the issue Quantum DLT2000 Tape Drive . . . . . . . . . . . . 20 Problems with DLT2000 on VMS V5.5-2 ALL-IN-1 Document Recovery . . . . . . . . . . . . 22 How to restore a clobbered drawer of documents Command Sequence in Windows NT . . . . . . . . . . 28 Execution sequence of a .BAT file Sendmail Secrets . . . . . . . . . . . . . . . . . 31 Unix mail systems and system load averages About the DECUServe Journal . . . . . . . . . . . 33 Contact Information . . . . . . . . . . . . . . . 34 The DECUServe Journal December, 1997 Page 2 From the Editors' Keyboard From the Editors' Keyboard ---- --- -------- -------- Your faithful and devoted editors are pleased to close out 1997 with yet another issue of the DECUServe Journal. In preparing this edition, we were astonished to discover that we had actually managed to fall several months behind on one of our mainstay conferences, VMS. (The VMS Notes conference on DECUServe is the official home of the silent "Open".) So the theme for this month might be "round up the usual suspects." In addition to a number of tidbits from the VMS conference (we especially enjoyed the discussion of $ PIPE), we also pay visits to our friends in the ALL-IN-1, hardware, Unix, and NT arenas. Not a bad way to finish the old year and begin the new. And since this is a propitious time to reflect on things past and contemplate changes ahead, we thought we'd remind the active DECUServe subscribers among our readers of an old, lapsed practice: You may, you are in fact enthusiastically encouraged to, nominate notes topics for inclusion in the Journal. If you see something that you feel shows DECUServe at its best, by all means let us know! We wish our readers all the best for 1998, and many happy s. * * * * * The DECUServe Journal December, 1997 Page 3 DCL PIPE Magic DCL PIPE Magic --- ---- ----- Abstract: The DCL PIPE command, being relatively new (and culturally foreign) to many VMS folks, has some unexplored potential. DECUServe's crack DCL hackers are on the case, though. As so often happens, the following discussion yields more than one way to get the desired result ... and also a pointer to a very nifty utility! (Be sure to read that last reply, though!) Participants: John Briggs, Rob Brown, Kevin Carnes, Stephane Germain, Brian Schenkenberger. Conference: VMS Note 2874.0, 1-Oct-1997 Carnes: Can PIPE redirect output to a symbol? --------------------------------------------- Has anyone become a PIPE expert yet? I want to use the pipe command to redirect the output of a command and somehow store that output in a symbol (for use in a subsequent command procedure). I have a program called ti that prints out a one-line title. I've tried all of the easy things I could think of like: $pipe ti |symb :== or $pipe ti |read sys$input symb all to no avail. I can do it by redirecting output to a temporary file, but that means there is a file to clean up every time. Does anyone know how to do this, with or without the PIPE command? Note 2874.1, 1-Oct-1997 Schenkenberger: SYMBOL can do it -------------------------------- Correct me if I'm wrong but I believe the PIPE implements the chaining of the various command elements by executing them in subprocess context. I say this because the process requires, at a minimum, TMPMBX privilege. Therefore, because you'll be creating the symbol in the context of an- other process, you'll not see it in the process that invokes the PIPE command. The DECUServe Journal December, 1997 Page 4 DCL PIPE Magic However, if you get yourself a copy of my SYMBOL program, you can use it to accomplish the task at hand with the following command syntax: $ PIPE SHOW TIME | SYMBOL/SET/ID="''F$getjpi("","PID")'" XYZZY $ SHOW SYMBOL XYZZY XYZZY == "1-OCT-1997 15:49:55" or, better still,... $ SYMBOL XYZZY [-] XYZZY == "1-OCT-1997 15:49:55" $ Note 2874.2, 1-Oct-1997 Schenkenberger: PIPE _does_ use subprocs ---------------------------------------- I'm not wrong. The following was used to verify my suspicions. $ PIPE WAIT 0:10:00 | WAIT 0:10:00 Looking with a SHOW SYSTEM revealed two subprocesses of the master process. Note 2874.3, 1-Oct-1997 Carnes: Where do I get a copy? ------------------------------ OK, I'll bite. How do I get myself a copy of your SYMBOL program? Note 2874.4, 1-Oct-1997 Germain: Where to get SYMBOL and lots of freeware ------------------------------------------------- Take a look at www2.wku.edu/www/fileserv (ftp.wku.edu also works) for most VMS freeware stuff. Note 2874.5, 2-Oct-1997 Schenkenberger: SYMBOL V4.0 on DECUS spring 1997 collection ----------------------------------------------------------- Actually, your best bet is to get the latest from: ftp://ftp.montagar.com/decus/vmslt97a/tmesis/ The DECUServe Journal December, 1997 Page 5 DCL PIPE Magic because the wku archive version is 3 1/2 years old! This has the latest and greatest version of SYMBOL V4.0. It was placed on the Spring 1997 DECUS collection. Also available there is SNAPSHOT. This is a package loosely based on Nick DeSmith's PDUMP utility. SNAPSHOT works on both VAX and Alpha and clusterwide; PDUMP is VAX only. SYMBOL V4.0 has lots of new features thanks mainly to Jon Pinkley. He was the force behind the addition of the subprocedural symbol manipulation that V4.0 now contains. I also added a new technique for copying process memory from one process to another directly! No more going through an allocated chunk of nonpaged pool or, in this case, nonpaged ocean! Note 2874.6, 2-Oct-1997 Brown: Confused --------------- >$ PIPE SHOW TIME | SYMBOL/SET/ID="''F$getjpi("","PID")'" XYZZY ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ I know I'm off topic here, but it is not clear to me how this part has anything to do with the SHOW TIME command. Note 2874.7, 3-Oct-1997 Briggs: Simple -- if you're a Unix weenie ----------------------------------------- The command line in question is parsed as: PIPE (Interpret remainder of command line as PIPE command) SHOW TIME (Execute SHOW TIME command in a subprocess with output directed to a pipe) SYNBOL/SET/ID... (Execute SYMBOL command in a subprocess, taking input from the pipe) The "''F$getjpi("","PID")'" is substituted by the DCL parser, generating the PID of the main process. The SYMBOL program runs, takes the current time as a line read from SYS$INPUT (the pipe) and sets it as a DCL symbol in the context of the parent process. Note 2874.8, 3-Oct-1997 Schenkenberger: How SYMBOL works... ----------------------------------- A few things here you need to know. First, SYMBOL. SYMBOL is a package I wrote several years ago and I have enhanced it quite a bit since its first release. SYMBOL permits one to set, delete and display DCL symbol in any process in a VMScluster. The syntax for setting a global symbol in a process is: The DECUServe Journal December, 1997 Page 6 DCL PIPE Magic $ SYMBOL/SET/ID=

The original question was: "How to place a program's output into a DCL symbol?" In the PIPE command, I used SHOW TIME to generate a line of output. It will then be used as the in the SYMBOL/SET command. The /ID="''F$getjpi("","PID")'" is used to set the symbol XYZZY in the process issuing the PIPE command. DCL will evaluate lexical F$getjpi() in the context of the process issuing the PIPE command. Thus, SYMBOL will be running in a subprocess of the process issuing the PIPE command and will set the symbol in the context of the PIPE command process. Sequence of events: 1. Issue PIPE command. 2. DCL parses command/evaluates lexicals. 3. Process spawned to execute SHOW TIME / SYMBOL... 4. PIPE takes output of SHOW TIME and passes as input to next subproc. 5. SYMBOL runs in subproc. Sets symbol XYZZY in master proc to value passed as input by PIPE. 6. Subprocs die. BTW, the 'SHOW TIME' could very well have been a SHOW SYSTEM. However, only the first line would make it into the XYZZY symbol. For example: $ PIPE SHOW SYSTEM | SYMBOL/SET/ID="''F$getjpi("","PID")'" XYZZY $ SYMBOL XYZZY [-] XYZZY == "OPENVMS V7.1 ON NODE ALPHA 3-OCT-1997 08:12:04.77 UPTIME 4 03:09: 52" $ Note 2874.9, 3-Oct-1997 Carnes: Thanks, it works. ------------------------- Thanks, I retrieved the symbol package and it works as advertised. I had hoped that there was a way to use plain old DCL to do this, but evidently not. Note 2874.10, 6-Oct-1997 Briggs: DCL output => DCL symbol -------------------------------- Fear not, gentle reader -- it can be done. $ pipe show time | ( read sys$input a ; define /job foobar &a ) $ dcl_symbol = f$trnlnm ( "foobar" ) The DECUServe Journal December, 1997 Page 7 DCL PIPE Magic The keys here are the observations that the semicolon (";") delimiter causes sequential execution within a single process and that the ampersand ("&") operator performs symbol substitution in subprocess context. And, of course, job logicals. One liner: $ pipe show time | ( read sys$input a ; define /job foobar &a ) ; dcl_symbol = f$trnlnm("foobar") Note 2874.11, 6-Oct-1997 Schenkenberger: Interesting... ------------------------------ You may want to add " ; deassign/job foobar" to the end of the "one liner" avoiding the %DCL-I-SUPERSEDE, previous value of FOOBAR has been superseded message if used for subsequent symbol definitions. Interesting. What is you run out of JTQUOTA? ;) Note 2874.12, 6-Oct-1997 Briggs: Production touch-ups ---------------------------- Nice touch. Probably slightly better than DEFINE /NOLOG. Another production touch-up would be /NOLOGICAL and /NOSYMBOL on the PIPE command. >Interesting. What is you run out of JTQUOTA? ;) Then the PIPE command itself is likely to fail with %SYSTEM-F-EXLNMQUOTA. A failure of the DEFINE /JOB command yields identical symptoms. If you run out of JTQUOTA, you should whine for more. ;) Note 2874.13, 8-Oct-1997 Schenkenberger: SYMBOL V4.0 Update/NOTICE ----------------------------------------- Potential system crash when using SYMBOL V4.0 on OpenVMS VAX V7.1 ----------------------------------------------------------------- *** NOTICE *** -------------- If you are using the SYMBOL V4.0 release (SYMBOL040-97143) which was distributed on the Spring 1997 DECUS OpenVMS collection and you have plans to install it on or up upgrade to OpenVMS VAX V7.1, you should contact me at VAXman@TMESIS.COM for an update for SYMBOL which fixes a problem under OpenVMS VAX V7.1. PROBLEM DESCRIPTION The DECUServe Journal December, 1997 Page 8 DCL PIPE Magic ------------------- Digital increased the maximum of the SYSGEN parameter CLISYMTBL from 500 pages to 1024 pages. If you are use the SYMBOL040-97143 release and increase this parameter above 500 pages, your system will crash. This update increases the maximum symbol table size that SYMBOL will handle to this new CLISYMTBL parameter maximum of 1024. About AMDS ----- ---- Abstract: Ordinarily, and for obvious reasons, we generally don't consider for publication in the Journal discussion threads that do not reach a conclusion. We're making an exception in this case; although no real closure is reached on the main topic (though non-VMS network problems are hinted at), the sidebar discussion of DECamds may be of interest to some of you VMS system managers on tight budgets -- and who isn't, these days? Participants: Alan Bruns, David Campen, Linwood Ferguson, Terry Kennedy, Bart Lederman, Keith Parris, Gary Rice. Conference: VMS Note 2871.0, 17-Sep-1997 Rice: Current VMS Performance Tools? ------------------------------------ We are experiencing intermittent problems with our environment. The environment is fairly complex, thus posing a problem for me in deciding just which conference to post this question in. So, please forgive me is it seems like I picked the wrong one. Having said that, let me try to describe the environment: We have a large network with 100s of machines connected. Most of them are Macs and Vaxes. On 1 segment of the net (isolated from the others via a router) is a small collection of Macs and Vaxes that "talk" to each other constantly via some applications that we wrote. One Mac acts as a focus for a total of 16 network mailboxes (DECnet) connected to a total of 4 vaxes (that's 4 mailboxes on each vax). The software exchanges packets (of varying lengths) between the Mac and each vax mailbox at about 2 packets /second. On a second Mac, 16 LAT sessions are in progress. Again, there are 4 The DECUServe Journal December, 1997 Page 9 About AMDS sessions hosted on each of the 4 vaxes. Now the problem: at random times, users of the system report "sluggish response times". This is happening at a rate of roughly 5 "events" a week, based on 4 users running on a 7 by 24 schedule. The users have found that by logging out of a single LAT session and re-initiating it, the problem can be cleared. We have seen some problems on the net (using a Network General Sniffer), but they don't seem to be directly related to the performance issue. So, with that background, my question: Is there some sort of VMS-based performance tool that is available (either commercially of PD) that could help me ELIMINATE or CONFIRM the VAX as the cause of the slowdown? Note that any tool suggested needs to work with VMS v6.1. Note 2871.1, 17-Sep-1997 Campen: Network sniffer should help. ------------------------------------ It would seem that the network sniffer would be of some help in localizing the problem. Presumably the interactive response depends on a network packet leaving one machine and then a response returning over the network from another machine. If the network sniffer shows a long interval between the request and response packets; then that would indicate that the problem is with the responding machine and a short interval would indicate that there is no problem with the responding machine. Note 2871.2, 18-Sep-1997 Lederman: Try DECamds. ---------------------- If you haven't already, please look at DECamds. It's on the VMS distribution kit, but has to be installed separately. It can be very useful in tracking down when and where problems happen. You can litterally point-and-click on system or problem area to zoom in on the statistics. You can also have it monitor a system and send out an alarm when resource usage exceeds a threshold. We find it very useful here within Digital, but apparently it's not as well known as it should be in the user community. It used to require a cluster or workstation license to run, but it sounds like you have that. If not, get a 7.1 distribution kit: AMDS now only requires any VMS license. The newer versions of AMDS generally will install on older versions of VMS, because it's best to run the same version on all nodes in a network, and Digital realizes that people run mixed environments (we do too). The DECUServe Journal December, 1997 Page 10 About AMDS Note 2871.3, 18-Sep-1997 Bruns: By all means, run the 7.1 version of AMDS ------------------------------------------------ For sure get the 7.1 version - I was never able to get "FIX" commands to work under the versions shipped with prior VMS releases. We have a DECforms application that occasionally goes into a loop - it's really easy to drill down on the process doing that (two double-clicks, actually) and then do a forcex on the process (or a stop). I keep the AMDS system overview window up on my workstation all the time. Good stuff. We're running VMS V6.2, by the way. Note 2871.4, 18-Sep-1997 Ferguson: And it FIXES stuff is the best part --------------------------------------------- > We find it very useful here within Digital, but apparently it's not as > well known as it should be in the user community. Sadly so. I think this is the ONLY evidence that Digital Marketing is effective. Since this is not sold, they don't market it, and it does seem _slightly_ less well known than stuff they do market. I was delighted a couple of months ago when I found it could adjust quorum on a running cluster that someone accidentally added a votes=1 workstation to and then removed it, and then we tried to take down one other node and locked up tight. 2 minutes and AMDS and it was happy. (1 minute 50 to find the right place, 10 seconds to fix). It does lack performance history information, and hot file information, but it is one dynamite tool for a freebie. Note 2871.5, 18-Sep-1997 Kennedy: There's a benefit here 8-) ----------------------------------- > Sadly so. I think this is the ONLY evidence that Digital Marketing is > effective. Since this is not sold, they don't market it, and it does > seem _slightly_ less well known than stuff they do market. Look at the bright side - this probably means it won't be sold off as soon as you start needing it. Note 2871.6, 18-Sep-1997 Parris: -------- >Now the problem: at random times, users of the system report >"sluggish response times". This is happening at a rate of roughly 5 >"events" a week, based on 4 users running on a 7 by 24 schedule. In addition to the suggestion in .1 of looking at request/response latencies, The DECUServe Journal December, 1997 Page 11 About AMDS I'd want to monitor for lost, corrupted (and then retransmitted) packets, and for overall traffic levels, by protocol type if possible (to see if there's any correlation between the errors and a period of time during which the traffic is noticeably different in either quantity or protocol mix). >The users have found that by logging out of a single LAT session >and re-initiating it, the problem can be cleared. Does this potentially connect them to a different system offering the same LAT service? >We have seen some problems on the net (using a Network General >Sniffer), but they don't seem to be directly related to the performance >issue. Could you describe the problems that you have seen? Note 2871.7, 19-Sep-1997 Rice: ------ > Does this potentially connect them to a different system offering the same > LAT service? Potentially yes, but ONLY if the primary CPU failed at the same time that the session was restarted. > Could you describe the problems that you have seen? Sort of. I didn't actually see the sniffer info. It was reported to me (and others) by the network engineer. He described the problems as "long ACK times" (on the order of 2 seconds) associated with DECnet packets. We have almost concluded that the VAX (sysgen, uaf, tuning, etc.) is not at issue and are pursuing this as a network problem (probably hardware related). When Faster Networks Don't Help ---- ------ -------- ----- ---- Abstract: Following are some interesting observations on high-speed LAN configurations. Readers are cautioned that comfortable certainties may be replaced by nagging doubts after reading this article. Moral of the story: Test, test, test. The DECUServe Journal December, 1997 Page 12 When Faster Networks Don't Help Participants: Jack Harvey, Alan Hunt, Keith Parris, Brian Schenkenberger. Conference: VMS Note 2868.0, 27-Aug-1997 Hunt: Comments on this configuration? ------------------------------------- Does anyone have any comments on the idea of a VMScluster mixing three large VAXs, 20+ VAXstations, Turbolaser, and 18 Alphastations using a mixture of 10 MB and 100 MB ethernet? The VAXs would be on 10 MB while the Alphas would be on 100 MB. Presently the three VAXs and 38 VAXstations are living alone on a 10 MB network that is already over loaded. The VAXstations are model 4000/90s. They are proposing switched ethernet but the big VAXs are already on switched ethernet. Note 2868.1, 28-Aug-1997 Parris: -------- A general answer could be "it will probably work just fine", but your level of satisfaction probably depends on how you plan to use the systems in the cluster. If the VAX/VAXstation group and the TurboLaser/Alphastation group are just being grouped together for reasons such as ease of system management, centralized backups, making things easier temporarily while you migrate from one architecture to the other, or for ease of occasional file transfers, and you don't plan to heavily access a common set of shared files all the time (and thus have to contend for locks on those resources), you'll probably do about as well with the systems together as separated. How sensitive are the users and the applications to pauses (such as one sees in conjunction with cluster state transitions)? With an overloaded LAN, folks tend to run with higher-than-default values of RECNXINTERVAL to avoid CLUEXITs, causing the impact of something like someone tripping over the power cord to a satellite node greater than with the default value. And the more nodes you place in a cluster, the higher the probability one will get halted or powered off accidentally. Perhaps you could expand on your particular situation, and the alternatives you are currently considering. The DECUServe Journal December, 1997 Page 13 When Faster Networks Don't Help Note 2868.2, 28-Aug-1997 Hunt: More info --------------- The VAXs (non-workstation) would still have up to 200-250 users on at one time (1400-1700 accounts) and the disk storage/tape backup would be shared among all systems. There would be four servers (3 VAX 7630s and 1 two-cpu Turbolaser) serving up to 100-130 disk drives (48 RA92, 8 RA90, ~72 RZ28 on 4 HSC70s and 6 HSJ40s) to roughly 40-50 workstations. Each server has 2 CI ports. I have a rough time with a VAX connected to 10 MB trying to feed data to 100 MB Alphastation 500/333. The Alphaserver is being purchased with the 625 Mhz upgrade. Presently the ethernet in use has seen steady state 50-60% usage with spikes I believe up to the 80-90% level. Session pauses (editor freezes) are common. We have a modeling application (models the engine/transmission) that runs on the VAX systems now. The Alpha hardware is being purchased to run just the model application. The model consists of Fortran modules to model the components of the engine/transmission plus the software program that controls the powertrain and runs on the microprocessor in each vehicle. Much of this information comes from users on the VAX systems. Over time I see other apps moving to the Alpha as well and this being one big time sharing/server cluster. Our networking guys are saying that the SWITCHED ethernet (10 and 100) they use from BAY Networks will provide better throughput than FDDI. The servers are in one computer room but the workstations are spread through a building and connected to concentrators via twisted pair. The concentrators are connected by fiber. The BAY equipment in use now are models 350F and 350T. I have a problem with having much of this on a 10 MB network. Thanks for the feedback. Note 2868.3, 29-Aug-1997 Harvey: Ya gotta have faith... ------------------------------ > Our networking guys are saying that the SWITCHED ethernet (10 and 100) > they use from BAY Networks will provide better throughput than FDDI. A bunch of our non-production nodes were just moved from a conventional (seriously overloaded) 10 MB Ethernet segment to an Etherswitch. I measured the transfer rate using DTSEND before and after the move. And I was fairly careful about the measurements. The rate was measured between an Alpha 2100 and twenty different VAX nodes of various types. Each pairing was measured three times, one minute runs using 2000 byte packets. Overall, the performance after the Etherswitch was installed was worse by about six percent. Many node pairs were degraded by ten percent and some by about 25 percent. Only four were improved. The DECUServe Journal December, 1997 Page 14 When Faster Networks Don't Help How come? Well, for one thing, these measurements were made at hours when the network was relatively quiet. But they demonstrate a point: It isn't easy to show a network performance improvement. So why didn't I measure when the network was busy (before and after) to get a realistic test? I did, but the data was so highly variable from trial to trial that I didn't have a statistically believable result. Note 2868.4, 30-Aug-1997 Hunt: Did the switches seem better? ----------------------------------- Did the systems seem to be better or worse from the user's standpoint? One other issue here is that multiple switches will be need to connected together since our systems are distributed inside the building. Probably 6-7 switches at a minimum. I wouldn't expect any improvement between systems on different switches. Thanks for the feedback. Note 2868.5, 31-Aug-1997 Parris: -------- Re: .2 > I have a rough time with a VAX connected to 10 MB trying to feed data > to 100 MB Alphastation 500/333. Me too. > Presently the ethernet in use has seen > steady state 50-60% usage with spikes I believe up to the > 80-90% level. Session pauses (editor freezes) are common. ..probably a symptom of lost packets which must be retransmitted. > We have a modeling application (models the engine/transmission) that > runs on the VAX systems now. The Alpha hardware is being purchased to > run just the model application. The model consists of Fortran modules Could the storage for just the modeling application be moved to the Alphas, and served back to the VAXes? or a shadow member put on the Alphas, so they could read from their local copy? > Our networking guys are saying that the SWITCHED ethernet (10 and 100) > they use from BAY Networks will provide better throughput than FDDI. Most switching bridges behave much differently than the LAN equipment that was present when PEDRIVER's assumptions were made. As long as they act strictly The DECUServe Journal December, 1997 Page 15 When Faster Networks Don't Help like bridges, you're probably OK, but most are designed for the IP market, where I fear there are assumptions like "it's OK for the switch to throw away multicast packets when things get busy -- it's probably just an ARP broadcast storm anyway" which could wreak havoc with PEDRIVER's Hello multicasts. If it doesn't work out, it would be nice if the network folks allowed you the altenative of a GIGAswitch and FDDI links. Note 2868.6, 3-Sep-1997 Hunt: Bay Networks white paper ------------------------------ The "chosen" network equipment provider is Bay Networks. On their web site (www.baynetworks.com) there is a white paper comparing switched 100 Mb Ethernet versus FDDI (/Federal/need.html). Here is a quote, "... as new devices are attached to the FDDI backbone ring, the effective bandwidth to each device decreases exponentially." This is totally the opposite of anything I have ever read/heard before. In fact, FDDI uses a "token" and adding additional nodes should cause it to degrade linearly. Any comments ? Note 2868.7, 4-Sep-1997 Parris: -------- As is the case with any shared bus, if there is one link between nodes, it will get all the bandwidth. For two links, each will get 1/2. Basically, for "n" links, each link gets a fraction 1/n of the bandwidth. I don't think I'd describe this as an exponential curve, but you're reading marketing literature, right? And you're right: because FDDI uses a token-passing mechanism, it's arguably more stable under near-saturation conditions than a Collision-Sense-Multiple-Access (CSMA) bus like Ethernet. But ask them if they have a non-blocking crossbar switch for their backplane, like the GIGAswitch, or if they simply use some sort of shared bus. If there's a shared bus, its bandwidth degradation will have an identically-shaped curve to the one they're describing as "exponential", of course. :-) Note 2868.8, 5-Sep-1997 Schenkenberger: ---------------- >Collision-Sense-Multiple-Access (CSMA) bus like Ethernet. Carrier Sense, Multiple Access 'collision' comes in with CD -- Collision Detect. The DECUServe Journal December, 1997 Page 16 When Faster Networks Don't Help Note 2868.9, 7-Sep-1997 Harvey: No visible change ------------------------- > Did the systems seem to be better or worse from the user's standpoint? Depends on their mood when you ask them. :-) I learned long ago that asking users if a change helped yields random results. I also do a lot of disk-to-disk backups in parallel on this LAN. No noticeable change, faster or slower. DECWindows Won't Start for User ---------- ----- ----- --- ---- Abstract: It's the little things that'll get ya. The fact that small overlooked details can have large consequences is an inherent characteristic of complex systems. Wait ... isn't that very similar to chaos theory? Hmmmmm. Not that we would want to suggest that any MIT-derived windowing system is chaotic.... Participants: Ken Akerboom, Rob Brown, Lee Gleason. Conference: VMS Note 2875.0, 2-Oct-1997 Gleason: DECwindows access by non-SYSTEM accounts ------------------------------------------------- Last week, I went to a very remote job site, to a refinery, to work with a group that will be using an AlphaStation 255/233 for plant control. My task is to install the software, get it onto the ethernet, talking to the plant and assorted PC's involved. At first, the DEC side of things went pretty well. The Alphastation came with OpenVMS V6.2-1H3 from the factory - good news there. I installed UCX V4.1-2, UCXECO_41_2 (I figured, if a UCX eco was on the distribution CDs, it must be good), MMOV 2.0 (some sort of multi media package, from DEC), and OPEN3D V3.6 (from DEC as well). It's running DECWindows Motif V1.2-4. I also configured DECnet. I followed all directions in the kits for checking & changing SYSGEN params and account quotas - they were all pretty reasonable sounding. I thought I was home free, but, I created some new user accounts, The DECUServe Journal December, 1997 Page 17 DECWindows Won't Start for User and tried using them to make sure they work OK. Apparently, I have some sort of DECwindows problem here.... I could log in and out at the console graphics display as many times as I wanted, using New Desktop or the older DECwindows interface. I could log in to the console one time using any other account, and then log out, no problem...but, if I then attempt to log in again, at the console, with that same account, I see the New Desktop hourglass flash by, then the screen goes to a blank grey basket weave background, and an arrow mouse cursor. The cursor will move, but nothing else happens, regardless of what keystrokes or mouse clicks you try. I didn't have any other way to log in at that point, so I had to reboot each time. I could login using one of these new accounts, then logout, and then login in and out N times with the system account, and, then, try the user account again, and the grey screen reappeared. I created a test account with all of the privs of the SYSTEM account, and with the same UIC [1,4]...problem still occurs... A look at the DECWindows logs shows a clue...the DECW$SM.LOG for the user accounts will have a final line that reads XIO: unable to open connectio WSA1: after 0 requests (0 known processes) with 0 events remaining. Exiting XSESSION.COM SYS$MANAGER:DECW$SERVER_0_ERROR.LOG ended with 1-OCT-1997 12:45:44.6 Calling the dispatcher... 1-OCT-1997 12:45:44.7 Invalid access from transport: LOCAL node: 0 user: GLEASON Looks like the DECwindows server has some memory of the logins of non-SYSTEM accounts, and blows up because of it. I'll be going back there next week. Anyone else ever seen this problem? Know of a fix or patch? Have any advice at all? I don't want to stay at the refinery forever...this is no garden spot... on top of that, it has no phone access to speak of, so I can't call support from there... Note 2875.1, 3-Oct-1997 Akerboom: Password expired? --------------------------- I'd guess the password was "pre-expired" and the set password dialog is hosed for some reason. Test by logging in on SYSTEM account. Create a "virgin" new user account. Do a "set host 0" (assuming DECnet is available, else use telnet) and login once. Log out and log in again using set The DECUServe Journal December, 1997 Page 18 DECWindows Won't Start for User host/telnet. Any messages? This assumes you have sufficient log-ons available. Note 2875.2, 3-Oct-1997 Brown: ------- Good point. Also check for the account being DISUSERed. That's what usually catches me. Note 2875.3, 3-Oct-1997 Akerboom: A couple more stabs in the dark ----------------------------------------- Another thought occurred to me... Make sure that there is a window manager in the "automatic startup" group ("Session manager"->"Options"). Although if you get it once, it should be there the second login. Another thought... Make sure the user account has write access to their login directory (or wherever you point their DECW$User_Defaults logical). If the system expects to find something there, it may get confused if there isn't anything. It is conceivable that for the 1st login, all the data is sucked from the "system" area, so all works ok, but then the second login it gets confused if nothing for the user exists). Note 2875.4, 28-Oct-1997 Gleason: it was LOGOUT symnbol redefined... ------------------------------------------- Just returned from that remote site...it was a harrowing experience... local driver almost drove us into a tank of hydrofluoric acid, almost got tricked into sampling some local delicacies, such as Chinicuiles (Agave worms) and Escamoles (ant's eggs), and am suffering from a case of turista that is causing more pain than most motocycle crashes I've had... ...but, the answer to the mysterious lock up problem was traced to a redefinition of the LOGOUT command in the user's LOGIN.COM. Apparently, the CDE/MOTIC/DECwindows subsystems wil not work if the user redefines LOGOUT to be a COM file, and that COM file is missing... Thanks to everyone for the suggestions - they got me thinking about what could be going on there to cause the display to become unavailable. The DECUServe Journal December, 1997 Page 19 MicroVAX Terminal Ports MicroVAX Terminal Ports -------- -------- ----- Abstract: A short sequence of notes concerning some odd behavior on the part of a MicroVAX 3100's terminal ports, and a straight-to-the-hardware fix. (Isn't that worth a point on the Hacker Purity Test?) Participants: Elli Barasch, Linwood Ferguson, Marc Kozam. Conference: VMS Note 2870.0, 16-Sep-1997 Barasch: Funky terminal port behavior... ---------------------------------------- We had an odd discovery of behavior of setting modem characteristics on the TTA2 (modem controllable) port on uVAX 3100 family systems. These systems were running VMS 5.4-3 or 5.5-2. If you programmatically assert the DTR signal on a port, via $QIO(IO$M_SET_MODEM|IO$M_MAINT|IO$_SETMODE), it is unpredictable as to the state of the signal once the process exited. IO$_SETMODE is the 'temporary' characteristic option. We did this test on several 3140's and 3120's. There was no correlation to h/w or operating system version. In most, the signal stayed high, in some, it went low. Okay, so why should I care? Well, I upgraded a program that used to assert DTR with one that mistakenly forgot to. The application talks over the port to another system using a proprietary protocol - the partner's system has our DTR tied to its DSR. Without DSR asserted, the partner app sometimes dies a nasty death, or simply doesn't work. After the upgrade, everything worked fine (for weeks) until the next time we rebooted the system. The interface broke. What we think happened was that DTR remained set after the upgrade, and so the bug was suppressed. After the reboot, (and a system reset), the DTR was cleared, and the bug surfaced. BTW, this app runs on about 100 different nodes in our network - each connected to customer machines over which we have little control. Shouldn't a terminal port stop asserting any type of signal after the owning process exits (so long as it is a temporary characteristic setting)? Is there any way to sense the DTR state? What is the expected behavior? The DECUServe Journal December, 1997 Page 20 MicroVAX Terminal Ports Note 2870.1, 16-Sep-1997 Ferguson: What are the permanent characteristics? ------------------------------------------------- What are the port's permanent characteristics? If it's available for login, I thought DTR cycled on a 30 second basis or something like that. There's a flow chart somewhere (or was) that described how the modem signals were managed on an idle port. Note 2870.2, 17-Sep-1997 Kozam: Going to the hardware is a sure fix ------------------------------------------ I've log ago stopped attempting to get VMS to make these limited modem control ports behave in a reasonable (to me at least) fashion. My "fix" is to have a detached process that twiddles the hardware bits and thus controls the state of DTR. Note 2870.3, 25-Sep-1997 Barasch: Sure fix it is... -------------------------- .1: It's not available for login - (NOTYPEAHEAD is permanent, TYPEAHEAD is temporary) .2: That's exactly how we're dealing with this. We wrote a program to turn off the bits upon exit of the app. Thanks. Quantum DLT2000 Tape Drive ------- ------- ---- ----- Abstract: The following exchange covers a couple of matters relating to a Quantum DLT2000 tape drive on a VMS system -- questions, answers, and (please note) a positive experience with Digital's support lines. The DECUServe Journal December, 1997 Page 21 Quantum DLT2000 Tape Drive Participants: David Campen, Jim Campobello, Linwood Ferguson. Conference: HARDWARE_HELP Note 2177.0, 14-Nov-1997 Campobello: Quantum DLT2000 on VMS 5.5-2H4 problems --------------------------------------------------- We have a MicroVAX 3100 Model 90 running VMS 5.5-2H4. This morning we tried to replace an old Exabyte 8200 8mm tape drive with a new Quantum DLC 2000XT (single unit) tape drive. This resulted in two problems: 1. VMS doesn't seem to be able to handle the drive. I get a fatal controller error when I try to initialize the tape, and indefinite hanging (freezing the whole system) when I try to mount/for a tape. 2. The tape won't unload. The power-up light sequence looks fine, but when we push the "unload" button the "Tape in use" light keeps blinking, and all the right-side lights flash momentarily, and the tape never unloads. The handle remains locked and we can't get the tape out. Any suggestions would be mightily appreciated. Thanks. Note 2177.1, 14-Nov-1997 Ferguson: Probably lost leader ------------------------------ > 2. The tape won't unload. The power-up light sequence > looks fine, but when we push the "unload" button the > "Tape in use" light keeps blinking, and all the right-side > lights flash momentarily, and the tape never unloads. > The handle remains locked and we can't get the tape out. Everytime I've seen that was because the leader had come off inside the unit. It's probably time for a service call. It can be done if you fiddle with it a bit, there is a release for the tape on the inside on the right (I think) side, remove it slowly to ensure that the tape has not been extracted from the cartridge. If you remove the plastic case from the top of the mechanism (on the inside), the leader will be loose in there. It needs to be threaded around the outside of the exposed slot, and then hooked over the plastic do-thinggy that stuffs it into the cartridge. But if it's new, the by-far best way is to get service to fix it so you don't void your warranty or somehow break something. PS. I suspect there are other possible causes for the symptoms as well, just haven't seen them. The DECUServe Journal December, 1997 Page 22 Quantum DLT2000 Tape Drive Note 2177.2, 17-Nov-1997 Campobello: Got the answers --------------------------- Our operations guy called the vendor and was talked through extracting the tape. The drive came through fine, but that tape was ruined. In regard to the other issue, DEC Support had the answer on how to get VMS to control the drive. I downloaded and applied patch VAXSCSI03_061. That fixed the controller problem and the drive is working fine. BTW, my experience with DEC Support was a pleasant surprise. After going through the menu system, I was connected with someone right away (no time on hold), I was called back in about 45 minutes, and then called back again 30 minutes later for a conference call with a more expert person who had the answer. I was quite pleased with the process. Note 2177.3 Campobello: To compress or not? ------------------------------- I'm trying to decide if we should use hardware compression on the DLT tapes. I don't really need it to fit the data we have, so I wasn't going to use it. But then I read in the manual that data transfer is 2 to 3 times faster with compression on. The time element isn't really important, since all our backups run overnight and have plenty of time to finish. OTOH, if speed is the only issue, why not compress so that the occasional restore runs a little faster? Should we compress or not? Is data transfer really that much faster with hardware compression on? Is that true in both directions (writes and reads)? Is reliabilty equal? Are there any other issues? We don't exchange tapes with other sites, so that's not a problem. Thanks. Note 2177.4, 17-Nov-1997 Campen: Compression does improve the transfer rate. --------------------------------------------------- I use compression with TZ88 and TLZ09 drives because of the improved transfer rate. I have only measured the transfer rates during saves, I would assume that you would get the same benefit during restores. The DECUServe Journal December, 1997 Page 23 ALL-IN-1 Document Recovery ALL-IN-1 Document Recovery -------- -------- -------- Abstract: It is a well-known technical fact that things tend to break when those who know how to fix them are away. DECUS Symposia (as they are once again known) are no exception to this iron rule of nature, and so we find a subscriber in need of some help with restoring some deleted ALL-IN-1 documents. And where better to ask than here? Participants: Roger Bruner, Lynda Peach, Graham Pye, Don Vickers. Conference: ALL-IN-1 Note 1243.0, 4-Nov-1997 Bruner: Recovering Deleted "Stuff" ---------------------------------- While my boss is at DECUS event, it has fallen my lot to restore a deleted folder or drawer. All I have to go on is a six-step technical article that starts with "Create a temporary VMS account and an ALL-IN-1 account" and goes through to "Run TRU on the original account." My boss used this to restore something for me the other week, and I know he put something in the MANAGER account instead of using a temporary account. Any suggestins on what he might have done and how I should proceed? I'd be most grateful not to have to bother him while he's away. Note 1243.1, 4-Nov-1997 Vickers: A few resorts to visit ------------------------------- One would ASSUME that the manager person might have created a new drawer in the MANAGER account. The first thing I would do would be to get into the MANAGER account and look at the drawers in its file cabinet. One would assume the new drawer would be named something that would indicate it was being used for the recovery effort. You could use the FC DRM IAD option and specify ALLIN1 as the owning OpenVMS account to see all the drawers in the PARTITION that might have been created as a second resort. The third resort would be to get an index of the highest few documents in the MAIN drawer. The recovered documents might have the original old dates but the actual document numbers would be high if they were placed into the MAIN (or any other ) drawer. The DECUServe Journal December, 1997 Page 24 ALL-IN-1 Document Recovery Note 1243.2, 4-Nov-1997 Bruner: Thanks, Don! -------------------- Don, as always, I believe you've given me what I need to get started. Will let you know...Thanks, my friend! Note 1243.3, 4-Nov-1997 Bruner: Pondering aloud...so to speak ------------------------------------- Don, found it in MANAGER all right...appears to be in a directory structure with typical A1 time-based filename...appears to be one for each previous restore done. Wonder how restore was done exactly as for creating the beginning directory? Did that make any sense? Note 1243.4, 5-Nov-1997 Peach: Steps to recover a deleted ALL-IN-1 account -------------------------------------------------- Don't know if this will help you or not -- but these are the steps I've used to recover an ALL-IN-1 account. Steps 1-3 deal with how to obtain the backup savesets .... 4) If the account has been deleted in its entirety, follow standard procedure in creating a new account. If necessary, obtain information about the account (USERNAME, UIC, /ACCOUNT, etc.) from Accounts Management. Normally, you would try to put the account back on the same disk with the same UIC as before but that is not a requirement. The user's ALL-IN-1 account should be created. Then delete everything out of the directories before pulling the files from the full backup. NOTE: THIS APPLIES ONLY TO THOSE ACCOUNTS THAT HAVE BEEN COMPLETELY DELETED FROM THE SYSTEM. 5) DO THE FULL BACKUP FIRST. $ SHOW DEVICE MKA500: If device status is ONLINE, it is available for use. If ALLOCATED or mounted allocated, someone else is using the device. $ MOUNT/FOREIGN MKA500 $ BACKUP/LOG/SELECT=[dir...]filename - MKA500:save_set_name.bck/SAVE_SET - device-where-files-go:[dir...]/OWNER=ORIGINAL 6) Restore each of the subsequent incremental backups, starting with the most recent. (Reference Section 10.16, p. 10-41) These savesets The DECUServe Journal December, 1997 Page 25 ALL-IN-1 Document Recovery are on the chosen scratch device. $ BACKUP/LOG/SELECT=[dir]*.*;* - SCRATCH:save_set_name.bck/SAVE_SET - device-where-files-go:[dir...]/OWNER=ORIGINAL 7) Verify the account via the But from TeamLinks a previous version of my attempts to > create drawer show one test document I created in that drawer (not in > real one, of course) but not the docs! What's wrong? This makes me think you have two drawers in play here. Is the one in the Manager account supposed to be the same one you are accessing from TeamLinks? The DECUServe Journal December, 1997 Page 26 ALL-IN-1 Document Recovery Given that the drawer in Manager was supposed to be a restoration of a previous drawer I would think that the best course of action would be to use the Move drawer to another user option (SM MFC MD MDR) to move the drawer to an account more suitable for the drawer clients. You may also wish to Edit the drawer access to allow the proper clients to have access. Another approach could be to overlay the drawer you are accessing from TeamLinks by the directory structure and files for the drawer in Manager account. As I am sure you realize, the drawers created by the clients get 'Z' names for their top level directory file. The complete drawer structure includes all the normal file cabinet things of DOCDB, DAF and document directories. The ACCESS.DAT file has ACL entries defining the client access to the drawer. Note 1243.7, 6-Nov-1997 Bruner: Since Then... --------------------- Don, sorry if I didn't make things clear... Let me update you from earlier rather than start over again. I created a new drawer in MANAGER. I made note of the directory name and structure and copied the drawer to be restored to that directory (i.e, at VMS level I did COPY from where restore put drawer structure to where Create of new drawer wanted it). I gave proper access to the original user and to my own personal account so I could test it (in this case, READ=Y). I added the new drawer to the partition. In TeamLinks I added the access to the new drawer and it shows up just fine. But now when I click on it I get this message: Cannot list file object...the list operation failed due to file cabinet service errors So that's where things stand. Tell me what else I need to tell you, and I'll be happy to do so. Thanks! Note 1243.8, 6-Nov-1997 Vickers: MIGHT be a dirty cache ------------------------------- You said the documents in the drawer could be accessed from the Manager account as I recall. Did you do this via ALL-IN-1 IOS? Have you tried to access the new drawer from another account via ALL-IN-1 IOS? Assuming that ALL-IN-1 IOS can access the documents, it seems likely that the drawer structure is proper. The DECUServe Journal December, 1997 Page 27 ALL-IN-1 Document Recovery It MIGHT be that the file cabinet server has cached the drawer as it looked before you did the copy operations. You could look at the OA$LOG:OAFC$SERVER_ERROR.LOG file to see if any errors were noted there. Note 1243.9, 6-Nov-1997 Bruner: Rescued by Don Again! :-) --------------------------------- Don, I will try from another A1 account and check the log file as suggested. Thanks! I knew you could help! Note 1243.10, 7-Nov-1997 Bruner: Solved! --------------- Don, though the log didn't tell me a doggoned thing, you thing you said really struck me...the word "cache"...decided just to close down and restart Teamlinks instead of simply closing and reopening the File Cabinet connection, and -- wow! -- suddenly the right stuff was there! Thanks again (and again to Lynda)! Note 1243.11, 11-Nov-1997 Pye: Why it's $1000 a message ----------------------------- RE .4 <<<< BTW, I read in ComputerWorld the other day that Digital charges <<<< $1000 to its internal users to recover *ONE* message that has been <<<< deleted from Exchange. No-- not a typo $1000. That's because the restore process on Exchange works like this: 1) Find *another* *identical* server box to the one where you lost the message. Same disc size, etc... 2) Install Exchange, and restore the whole message store (Up to 16Gbytes) 3) Find the deleted message and give it back to the user on the other system. With later versions of Exchange (V5.0?) and Outlook (can't remember the version) this becomes a bit easier, since you can tell the server to pretend it's deleted messages, and only really delete them when a Janitor job runs after a few days/weeks. Before that, you can undelete them yourself. The DECUServe Journal December, 1997 Page 28 ALL-IN-1 Document Recovery In another year or two, we old hands won't be able to teach these new products any more new tricks :-) Note 1243.12, 13-Nov-1997 Bruner: That explains it! ------------------------- Ah, that does explain it, Graham. Oh, you will always find something else to add to products...btw, I was quite interested the other day to see that the script I had given Users to reset their A1 mail counts and is still in use was something I wrote in 1986! Funny how the counts between A1 and TeamLinks get off sometimes...some of the other TeamLinks problems are less amusing...*S* Graham, did you attend symposium? If so, did you meet my boss, Jack Surma? Note 1243.13, 17-Nov-1997 Pye: Maybe next time.... ------------------------ <<<< Graham, did you attend symposium? No, not this time. I was due to go on vacation on Thursday, so I would have needed to leave LA Wednesday lunchtime... Note 1243.14, 21-Nov-1997 Bruner: Vacation first for sure! -------------------------------- Too bad! My boss went for the first time ever, and I just was hoping he'd gotten to meet you. But vacation is clearly the better choice! *S* Command Sequence in Windows NT ------- -------- -- ------- -- Abstract: The problem here is how to serialize execution of commands in a Windows .BAT file, so that steps to clean up after a setup procedure don't start running until the setup procedure itself is done. Not all questions have good answers. The DECUServe Journal December, 1997 Page 29 Command Sequence in Windows NT Participants: John Briggs, Bruce Burson, Bob Fedick, Erik Husby, George Merriman. Conference: MS_WINDOWS_OS Note 328.0, 7-Nov-1997 Burson: Synchronizing CMD commands ---------------------------------- I'm *real* new to NT, and I admit I haven't had time to do an exhaustive search in this conference, so please overlook if this has been answered already; just point me in the right direction. I've got to distribute a software package to our NT 4.x workstations, and the preferred method is to package everything up in a self-extracting zip that automatically runs SETUP when it expands. No problem there, except that SETUP leaves some files on disk that don't need to be there. So I repackaged the zip so that it runs DOIT.bat which has the following lines in it (wrapped for readability): setup if exist "%USERPROFILE%/Start Menu/Programs/Unwanted Folder" del "%USERPROFILE%/Start Menu/Programs/Unwanted Folder/Unwanted File" if exist "%USERPROFILE%/Start Menu/Programs/Unwanted Folder" rmdir "%USERPROFILE%/Start Menu/Programs/Unwanted Folder" The problem is that the "if exist" lines execute before SETUP is done. I see my CMD window pop up, display all three lines and then disappear while SETUP is still doing its thing. SETUP created the unwanted folder and file, and by the time it's done the .bat file has long since finished and the unwanted files remain on disk. So my question is how can I tell the .bat file to wait for setup to finish before going on to the next lines? I don't want to put in a PAUSE because I want this to run without user intervention if possible. And I don't see a WAIT command. Any help will be appreciated. Note 328.1, 7-Nov-1997 Fedick: Errorlevel trapping, perhaps... --------------------------------------- Does the setup routine return an errorlevel upon completion? If so you could trap for that before continuing. It sounds like the .bat file is spawning a process for setup and I guess it wouldn't matter if setup returned an errorlevel as the bat file isn't in control any more. Good one... The DECUServe Journal December, 1997 Page 30 Command Sequence in Windows NT Note 328.2, 10-Nov-1997 Briggs: .BAT files are already sequential ----------------------------------------- I think you're out of luck on this one. My guess is that setup.exe forks off another process to do its work and returns immediately thereafter. > So my question is how can I tell the .bat file to wait for setup to > finish before going on to the next lines? I don't want to put in a > PAUSE because I want this to run without user intervention if possible. > And I don't see a WAIT command. Your .BAT file is already waiting for SETUP.EXE to complete. The problem is that SETUP.EXE isn't waiting for setup to finish. You could try to coerce the product installation script to do your deletion for you. There may or may not be such a script. Note 328.3, 10-Nov-1997 Merriman: This is the standard behavior --------------------------------------- > I think you're out of luck on this one. My guess is that setup.exe forks > off another process to do its work and returns immediately thereafter. The NT console command line processor always launches GUI applications in separate processes without waiting for them to finish. This is a legacy of 16 bit Windows behavior. Look into a product called WinBatch from Wilson WindowWare. (www.windowware.com) It knows about GUI windows (unlike the command line processor) and can, I think, solve your problem. Note 328.4, 10-Nov-1997 Husby: Try "start/wait" ----------------------- What about doing: $ start/wait setup That starts the setup exe in a separate process and then waits for it to complete before proceeding. The DECUServe Journal December, 1997 Page 31 Command Sequence in Windows NT Note 328.5, 13-Nov-1997 Burson: Separate user step - sigh --------------------------------- Thanks for the tips. Apparently my (and your) suspicion that setup calls something else to do the real work and exits immediately is correct. I've found no way to do what I want, so I split the "if exist" parts of my .bat file into a separate cleanup.bat and included that in the zip archive. The user's instructions have another step at the end that tells them to Start...Run... cleanup.bat. Sendmail Secrets -------- ------- Abstract: A brief exchange on one of many sendmail.cf arcana, namely how the system perceives and reacts to high system loads. Participants: Terry Kennedy, Pat Scopelliti. Conference: UNIX_OS Note 364.0, 24-Nov-1997 Scopelliti: Controlling sendmail -------------------------------- Need some sendmail help.... sendmail responds to system load average by refusing to accept connections if load averge is too high, or by skipping "lower priority" messages. This is evidenced by the sendmail queue getting large (over 3000!) and an invocation of sendmail -q -v quickly returning to the system prompt without delivering anything. e.g. # sendmail -q -v # <== appears within half a minute These trip points apparently are controlled by the config file sendmail.cf and the two settings Ox and OX. I want sendmail to always accept connections - I think I've got that done. But I can't seem to come up with a setting that means "always process the queue" no matter what load average is. Am I on the right path? The DECUServe Journal December, 1997 Page 32 Sendmail Secrets Note 364.1, 24-Nov-1997 Kennedy: --------- > Am I on the right path? Sort of. These two options control when sendmail refuses a connection and when it queues instead of delivering directly. The Oq config file option controls what gets done in queue runs - it is a factor that's applied to message priority using the formula: if msgpri > (q / (la - x + 1)) then deliver, else leave in queue where q is the setting of the Oq option, la is the current load average, and x is the setting of the Ox option. The message priorities are set else- where in the config file. However, all of this is for obsolete insecure sendmail versions - I'd suggest getting and building the latest version (currently 8.8.8). I'd suspect you're running on a Sun which has integer load averages (for example, 700 is a load of 7.00) which is skewing things. If not, try to see what is causing the high loading. Note 364.2, 25-Nov-1997 Scopelliti: Will Oq0 do it? --------------------------- The system is a Digital UNIX system (V3.2) which is part of our firewall. The system is heavily loaded - load averages can be 50 or more. Yes, it has decimal load averages, so 7.82 is possible. An upgrade is scheduled for December 15, but until then we've got to work with what we have. I just dont't want sendmail to queue files, so it seems if I set Oq to zero that it will not queue messages, right? Note 364.3, 25-Nov-1997 Kennedy: --------- Try it and see 8-) Note 364.4, 27-Nov-1997 Scopelliti: in process report ----------------------------- Looks like it's working... I have a job that once every half hour logs the number of mail messages in the queue (/var/spool/mqueue), the numer of sendmail processes running, and the number of deferred messages. The DECUServe Journal December, 1997 Page 33 Sendmail Secrets Whereas before when things got busy, we'd see the queues just grow and grow with the number of sendmail processes dropping, now I'm seeing the queue lengths staying 'reasonable' and the number of sendmail processes growing. About the DECUServe Journal --------------------------- The DECUServe Story (See http://eisner.decus.org/ for the complete story.) DECUServe is an electronic conferencing system, somewhat related to bulletin board systems but much larger and more organized. It is devoted to the general area of computer technology such as systems, software, hardware, and communication, in the Digital and related third party vendor market area. DECUServe also has complete access to and from the Internet. Usenet Newsgroups are accessible using newsreaders from DECUServe and the comp.os.vms newsgroup is added to a VAX Notes conference of its own. A DECUServe subscription is US$35 a year for both US and international residents. The conferencing system is available nearly 24 hours a day, seven days per week. There is no hourly connect charge. The subscriber pays communication costs to a phone number in eastern Massachusetts. Reduced rate communication services are available in some areas and INTERNET access is available (node - eisner.decus.org). Subscriptions must be used by a single person. Company or group subscriptions are not available, nor may subscriptions be transferred. DECUServe uses the Digital VAX Notes conferencing software. We currently have over 50 technical conferences available on subjects such as Security, the VMS Operating System, ALL-IN-1, Databases, Site Management, Personal Computing, DEC Networking, Third Party Software, Hardware, Workstations, the World Wide Web and many more. Over 130,000 technical notes are on line. All conferences, including the Frequently Asked Questions (FAQ) from the Usenet newgroups, are indexed to allow for fast text content searches. General information questions about DECUServe can be directed to the DECUS Customer Service line at 800-DECUS55 (800-332-8755). DECUServe applications are also available via E-mail. Send a message to Application@DECUServe.DECUS.Org. You may request a particular format by including any of the following in the mail The DECUServe Journal December, 1997 Page 34 About the DECUServe Journal subject; "ps", "postscript", "ln3" or "ln03". An ascii file will be sent if a format is not specified. Only one application per mail message will be sent. Publication Information Topic threads in the DEC Notes conferences on DECUServe are selected for publication on the basis of strong technical content and/or interest to a wide audience. They are submitted to the editor from various sources, including DECUServe Moderators, Executive Committee members, and other volunteers. Suggestions for inclusion are enthusiastically solicited. Articles selected for publication are edited on an OpenVMS VAX system in TPU and then formatted with Digital Standard Runoff. Contact Information ------------------- The editors of the DECUServe Journal are Brian and Sherrie McMahon. They can be reached by any of the following means: mcmahon_b@decuserve.decus.org mcmahon_s@decuserve.decus.org bmcmahon@cisco.com Work phone: +1 408 527 0434