http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user hourly 1 1901-01-01T00:00+00:00 http://gmane.org/img/gmane-25t.png http://gmane.org http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6285 <pre>Hi On a cluster running OGS/GE 2011.11p1 we have this configuration: a low priority queue low.q including all 28 exec hosts (1-28). a queue special.q including 9 exec hosts (9-17). an RSQ: limit hosts {*} to slots=$num_proc a threaded PE with allocation rule $pe_slots Threaded jobs require exclusive use of all 16 slots on a single exec host so when a threaded job is submitted to special.q, we want no low.q jobs scheduled on special.q's exec hosts (9-17) until the threaded job has a free host to run on. However low.q jobs should still be scheduled on exec hosts not included in special.q (1-8,18-28). Subordination or preemption seems like the answer but is there a way to configure something close to the behaviour we want? Thanks </pre> Ian Mortimer 2013-06-20T01:26:42 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6284 <pre>Hi Ryagz, Univa may be able to assist you with this. Their grid engine builds are well supported for the Windows platform and the cost is more than reasonable, especially for the level of support they offer. Hope this helps. -Ed -----Original Message----- From: Raghavendra [mailto:raghavendra-Wmf2PTAADboRmelmmXo44Q&lt; at &gt;public.gmane.org] Sent: Wednesday, June 19, 2013 06:05 PM To: users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org Subject: [gridengine users] Installing execution host on 64 bit windows 7machine Hi,Has anyone installed an execution host on a 64 bit windows 7 machine? If so can please provide with references to any documentation.I tried to install(OGS 2011.11) execution host through cygwin, but the installer was returning unsupported architecture. Please let me know wat else can be done.Thanks,Ryagz_______________________________________________users mailing listusers-GJWvCfsuOi7D8Lb4PCmq1TkpWUr6UEeh&lt; at &gt;public.gmane.org://gridengine.org/mailman/listinfo/users _________________________________________</pre> Ed Lauzier 2013-06-19T13:36:31 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6283 <pre>Hi, Has anyone installed an execution host on a 64 bit windows 7 machine? If so can please provide with references to any documentation. I tried to install(OGS 2011.11) execution host through cygwin, but the installer was returning unsupported architecture. Please let me know wat else can be done. Thanks, Ryagz </pre> Raghavendra 2013-06-19T22:05:39 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6282 <pre> On Jun 18, 2013, at 17:00 , Reuti wrote: Yes. If I have some time, I'll look at the sources. A. </pre> Esztermann, Ansgar 2013-06-19T08:34:56 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6281 <pre>Hi there, I was trying out an auto install of just the qmaster (so no execd hosts) using a deployment from the SGE 8.1.3 RPMs # cd /opt/sge # ./inst_sge -m -auto /opt/sge/util/install_modules/inst_vuwscifachpc02.conf and it seems to have worked, however, I noticed in the log file that there's this ---8&lt;---------------8&lt;---------------8&lt;---------------8&lt;------------ Adding default parallel environments (PE) ./inst_sge: line 881: SetupDefaultUsers: command not found starting sge_qmaster ---8&lt;---------------8&lt;---------------8&lt;---------------8&lt;------------ so went looking. If I grep through the installed files, I can't see any sign of it, other than the invocation itself ---8&lt;---------------8&lt;---------------8&lt;---------------8&lt;------------ #find . -type f -print | xargs grep -i SetupDefault ./inst_sge: SetupDefaultUsers $euid ---8&lt;---------------8&lt;---------------8&lt;---------------8&lt;------------ I tried a couple of other grep-s but have come up with nothing other than AddDefaultUsersets() which doesn't</pre> Kevin Buckley 2013-06-19T00:33:01 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6280 <pre>Hi Bill, I've created an area on github for now. Collaborators can request r/w access. I'll populate the area bi-weekly or when possible. Currently, just directory placeholders are there. https://github.com/elauzier/GridEngineContribScripts.git -Ed -----Original Message----- From: Bill Bryce [mailto:bbryce-38M+gKaC9ggAvxtiuMwx3w&lt; at &gt;public.gmane.org] Sent: Tuesday, June 18, 2013 09:58 AM To: 'Ed Lauzier' Cc: users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org, sge-discuss-VnOFtQRrI51aa/9Udqfwiw&lt; at &gt;public.gmane.org Subject: Re: [gridengine users] Script to query a running job's end time You could ask Dave Love if he has an area to put the contributed script examples. His version of Grid Engine is definitely more active from a development point of view than OGS. Regards, Bill. On 2013-06-17, at 5:02 PM, Ed Lauzier wrote: Hi, This brings up a good point. Do we have an area we can use for depositing/uploading user contributed script examples? ( I see a lot of scripts in many different areas....and sites...) Som</pre> Ed Lauzier 2013-06-18T16:03:49 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6279 <pre> Hi All. Apologies for the rather vague tone of this post, but I haven't got any detail on this yet... I've got a gridengine installation that's been running nicely for a couple of years now. It's a little troublesome in that we need to authenticate against a corprat Active Directory server, but we're not allowed to make any of the grid machines into member servers. I've worked around this by using sssd, which has been working very nicely for some time. The AD structure has just changed, with new users being added into a different domain than the existing ones. I've reconfigured sssd to deal with this, and things like getent work just fine. However - for reasons that I have not yet investigated - my calls to qsub and qrsh *sometimes* get a spurious "&lt; at &gt;domain" somewhere in the parameters (haven't found out where yet), which causes SGE to barf. I've printed out the string I'm passing to qrsh, and I can't see anything just yet... Can anyone give me any tips for how to track this down? Is there any hidden de</pre> Vic 2013-06-18T15:02:52 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6278 <pre>Am 18.06.2013 um 16:20 schrieb Esztermann, Ansgar: You are using RQS according to the above output? I observed it under some combinations of limits too. </pre> Reuti 2013-06-18T15:00:48 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6277 <pre>Hello List, recently, I've stumbled across a weird problem: jobs that should be eligible for scheduling (qalter -w p reports a "possible assignment") are not actually started. I see lines like this in spool/qmaster/messages: 06/18/2013 16:06:55|schedu|owl-master1|P|PROF: scheduled in 102.740 (u 115.210 + s 6.910 = 122.120): 0 sequential, 76 parallel, 4457 orders, 873 H, 227 Q, 834 QA, 1806 J(qw), 425 J(r), 0 J(s), 0 J(h), 0 J(e), 1 J(x), 4452 J(all), 63 C, 7 ACL, 173 PE, 58 U, 3 D, 0 PRJ, 1 ST, 0 CKPT, 0 RU, 1 gMes, 0 jMes, 4457/5 pre-send, -100/-172/-557 pe-alg Looking at common/schedule, I see STARTING lines for the same jobs repeated over and over (i.e. on each iteration). There is another weirdness involved here (I do not know to what extent it is related): node12-36 has gpu=0, so 151892 definitely should not be able to start there. BTW, the I've added gpu=0 after first noticing this problem with no gpu setting on the node at all. 151892:1:STARTING:1371564416:172860:P:openmp_fast:slots:8.000000 151892</pre> Esztermann, Ansgar 2013-06-18T14:20:40 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6276 <pre>You could ask Dave Love if he has an area to put the contributed script examples. His version of Grid Engine is definitely more active from a development point of view than OGS. Regards, Bill. On 2013-06-17, at 5:02 PM, Ed Lauzier wrote: _______________________________________________ users mailing list users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org https://gridengine.org/mailman/listinfo/users </pre> Bill Bryce 2013-06-18T13:58:14 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6275 <pre>I tend to keep my SGE configuration files (queues, user lists, ...) in a revision control system (I use subversion at the moment). Helps with things like that; have a branch per cell or so, and you can merge certain commits from one to the other. You could then have a cron job or so doing a checkout / config update. Should make it easy to apply certain changes to both branches, helps in case things go wrong ;) Also, a management system should be pretty good at things like 'if this file changed, do this' - we use cfengine (with the input files in subversion), and it certainly is. I know they seem work to implement, but it might be easier than using cron or so in the long run. For just a couple of files to copy and just a couple of SGE commands to run if they change, that should be quite straight forward to implement. Tina On 18/06/13 07:54, Reuti wrote: </pre> Tina Friedrich 2013-06-18T10:56:24 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6274 <pre>Dear Hugh and Reuti Thanks for your responses which have been very helpful. All the best Trevor The Scottish Association for Marine Science (SAMS) is registered in Scotland as a Company Limited by Guarantee (SC009292) and is a registered charity (9206). SAMS has an actively trading wholly owned subsidiary company: SAMS Research Services Ltd a Limited Company (SC224404). All Companies in the group are registered in Scotland and share a registered office at Scottish Marine Institute, Oban Argyll PA37 1QA. The content of this message may contain personal views which are not the views of SAMS unless specifically stated. Please note that all email traffic is monitored for purposes of security and spam filtering. As such individual emails may be examined in more detail. </pre> Trevor Carpenter 2013-06-18T09:18:40 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6273 <pre>Am 17.06.2013 um 23:51 schrieb Wagner, Justin: This individual distribution of settings have to be set uo by hand and a cron-job or alike. Especially as a defined complex might be needed to be attached to an exechost or queue (or removed from these first before the deletion). But in case you code the type of CPU in a complex, this can be different on various clusters then. This is nothing SGE can decide on its own what you would like to share ot not. </pre> Reuti 2013-06-18T06:54:27 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6272 <pre>The physical setup will not be identical. Generally I'd like to sync any configuration that is not tied to the actual physical setup. Specific configurations that I can think of right now are users, usersets, projects, departments, and consumable complex resources. Justin -----Original Message----- From: Reuti [mailto:reuti-HOsA5nJJ6nb33G8HivxkPnqWYbMAw+HU&lt; at &gt;public.gmane.org] Sent: Monday, June 17, 2013 2:35 PM To: Wagner, Justin Cc: users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org Subject: Re: [gridengine users] Syncing Grid Cell Configurations Am 17.06.2013 um 23:11 schrieb Wagner, Justin: Do all sites have the same physical setup, i.e. number of cores and machines/CPUs, projects, users, RQS, ... Otherwise the question would be: what subset would you like to sync? </pre> Wagner, Justin 2013-06-17T21:51:28 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6271 <pre>Am 17.06.2013 um 22:59 schrieb Wagner, Justin: As you want to influence scheduling, this also needs to be updated for each created project then - maybe in the sharetree. </pre> Reuti 2013-06-17T21:48:25 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6270 <pre>Am 17.06.2013 um 23:11 schrieb Wagner, Justin: Do all sites have the same physical setup, i.e. number of cores and machines/CPUs, projects, users, RQS, ... Otherwise the question would be: what subset would you like to sync? </pre> Reuti 2013-06-17T21:35:25 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6269 <pre>We have different physical sites that each have their own Grid setup as their own "cell". I've noticed that they keep completely separate configurations. This can be a hassle when there are certain configurations that I want to be synced across all those cells. Is there a way to have the configuration of one cell get pushed to another? I know I could accomplish this with a cron script to monitor and sync the two configurations, but I wanted to make sure I wasn't reinventing the wheel... Justin _______________________________________________ users mailing list users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org https://gridengine.org/mailman/listinfo/users </pre> Wagner, Justin 2013-06-17T21:11:43 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6268 <pre>Hi, This brings up a good point. Do we have an area we can use for depositing/uploading user contributed script examples? ( I see a lot of scripts in many different areas....and sites...) Someone could set up a free area on github if not already done. ( I can set up an area do this if not available. I see OGS has the sourceforge site...) Thoughts? Thanks, Ed -----Original Message----- From: Mike Hanby [mailto:mhanby-KQO9xKbDYp4&lt; at &gt;public.gmane.org] Sent: Monday, June 17, 2013 03:44 PM To: users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org Subject: [gridengine users] Script to query a running job's end time Howdy, I figured I'd share a script I wrote that is useful for querying running jobs to get their maximum end time (based on the h_rt). https://github.com/flakrat/misc-scripts/blob/master/query-job-endtime.rb The script can be run by passing it JobID's and/or UserID's For example: $ query-job-endtime.rb --userid johndoe,mickeymo --jobid 9189446,9189422 JobID TaskID Owner Max End Time Requested Ru</pre> Ed Lauzier 2013-06-17T21:02:35 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6267 <pre>Thanks for the quick response. The "-A &lt;account&gt;" would work for analysis only. I guess after thinking about it more, I want to keep the option of being able to influence the scheduling decisions based on the project. I guess I can write scripts that will keep the projects up to date in SGE as we define them in our other systems. Thanks, Justin -----Original Message----- From: Reuti [mailto:reuti-HOsA5nJJ6nb33G8HivxkPnqWYbMAw+HU&lt; at &gt;public.gmane.org] Sent: Monday, June 17, 2013 12:32 PM To: Wagner, Justin Cc: users-GJWvCfsuOi7Z+VzJOa5vwg&lt; at &gt;public.gmane.org Subject: Re: [gridengine users] Allow jobs to execute with non-configured Project Hi, Am 17.06.2013 um 20:37 schrieb Wagner, Justin: What about the "-A &lt;Account&gt;" option? It's an arbitrary string which can be used during job submission (defaults to "sge"), which is also recorded in the accounting file and available for `qacct` too. Even things like: $ qacct -j "*" | grep account | sort | uniq -c work. </pre> Wagner, Justin 2013-06-17T20:59:51 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6266 <pre>Hi, Am 17.06.2013 um 19:06 schrieb Trevor Carpenter: There is nothing special about such a setup. I install the qmaster and execd quite often on users' workstations (both are running on one and the same machine). Even with one core it would help to serialize your workflow and avoid overloading, but with many cores being available in a workstation nowadays, it makes even more sense. Depending whether you like to honor hyperthreading or not, it might be necessary to de- or increase the number of slots SGE detects during installation of the execd for the all.q queue, which is created for convenience during installation (check the "slots 8" or alike setting in `qconf -sq all.q`). Only hint for such a setup: I define "priority 19" in the all.q (which represents the nice value for jobs running in this queue), so that interactive work gets less impact by the jobs. You could also subtract one from the defined slot count and leaving it available for interactive access work - depends on your requirements. Unless</pre> Reuti 2013-06-17T19:45:54 http://permalink.gmane.org/gmane.comp.clustering.opengridengine.user/6265 <pre>Howdy, I figured I'd share a script I wrote that is useful for querying running jobs to get their maximum end time (based on the h_rt). https://github.com/flakrat/misc-scripts/blob/master/query-job-endtime.rb The script can be run by passing it JobID's and/or UserID's For example: $ query-job-endtime.rb --userid johndoe,mickeymo --jobid 9189446,9189422 JobID TaskID Owner Max End Time Requested Run Time ============ ======= ============== ===================== ============================== 9189446 0 jsmith 2013/07/29 06:40:05 41 Days 16 Hours 0 Mins 0 Secs 9189422 0 jblow1 2013/06/18 02:05:20 0 Days 12 Hours 0 Mins 0 Secs 9189242 0 johndoe 2013/06/25 13:03:40 8 Days 4 Hours 0 Mins 0 Secs 9189444 1 mickeymo 2013/06/18 00:39:20 0 Days 10 Hours 0 Mins 0 Secs 9189444 2 mickeymo 2013/06/18 00:39:20 0 Days 10 Hours 0 Mins 0 Secs 9189444 3 </pre> Mike Hanby 2013-06-17T19:44:27 Search the mailing list at Gmane query http://search.gmane.org/?group=$group=gmane.comp.clustering.opengridengine.user