Re: [lustre-discuss] Migrating files doesn't free space on the OST
Chad hit the nail on the head. I thought about the fact that it was still deactivated yesterday but was afraid to reactivate it until I verified the space was free. FWIW, the URL about handling full OSTs does not include the fact that the space will not be free until you reactivate the OST. It actually implies the opposite. http://wiki.lustre.org/Handling_Full_OSTs -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: Chad DeWitt Sent: Thursday, January 17, 2019 3:07 PM To: Jason Williams Cc: Alexander I Kulyavtsev; lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Migrating files doesn't free space on the OST Hi Jason, I do not know if this will help you or not, but I had a situation in 2.8.0 where an OST filled up and I marked it as disabled on the MDS: lctl dl | grep osc ...Grab the device_id of the full OST and then deactivate it... lctl --device device_id deactivate IIRC, this allowed the data to be read, but deletes were not processed. When I re-activated the OST, then the deletes were processed and space started clearing. I think you stated you had the OST deactivated. If you still do, try to reactive it. lctl --device device_id activate Once you reactivate the OST, the deletes will start processing within 10 - 30 seconds... Just use lfs df -h to watch... -cd Chad DeWitt, CISSP UNC Charlotte | ITS – University Research Computing 9201 University City Blvd. | Charlotte, NC 28223 ccdew...@uncc.edu<mailto:ccdew...@uncc.edu> | www.uncc.edu If you are not the intended recipient of this transmission or a person responsible for delivering it to the intended recipient, any disclosure, copying, distribution, or other use of any of the information in this transmission is strictly prohibited. If you have received this transmission in error, please notify me immediately by reply email or by telephone at 704-687-7802. Thank you. On Thu, Jan 17, 2019 at 2:38 PM Jason Williams mailto:jas...@jhu.edu>> wrote: Hello Alexander, Thank you for your reply. - We are not using zfs, it's an LDISKFS backing store, so no snapshots. - I have re-run lfs getstripe to make sure the file is indeed moving - I just looked for lfsck but I don't seem to have it. We are running 2.10.4 so I don't know what version that appeared in. - I will try to have a look into the jobstats and see what I can find, but I made sure the files I moved were not in use when I moved them. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: Alexander I Kulyavtsev mailto:a...@fnal.gov>> Sent: Thursday, January 17, 2019 12:56 PM To: Jason Williams; lustre-discuss@lists.lustre.org<mailto:lustre-discuss@lists.lustre.org> Subject: Re: Migrating files doesn't free space on the OST - you can re-run command to find files residing on ost to see if files are new or old. - zfs may have snapshots if you ever did snapshots; it takes space. - removing data or snapshots has some lag to release the blocks (tens of minutes) but I guess that is completed by now. - there are can be orphan objects on OST if you had crashes. On older lustre versions if the ost was emptied out you can mount underlying fs as ext4 or zfs; set mount to readonly and browse ost objects - you may see if there are some orphan objects left. On newer lustre releases you probably can run lfsck (lustre scanner). - to find what hosts / jobs currently writing to lustre you may enable lustre jobstats; clear counters and parse stats files in /proc . There was xltop tool on github for older versions of lustre not having implemented jobstats but it was not updated for a while. - depending on lustre version you have the implementation of lfs migrate is different. The older version copied file with other name to other ost, renamed files and removed old file. If migration done on file open for write by application the data will not be released until file closed (and data in new file are wrong). Recent implementation of migrate does swap of the file objects with file layout lock taken. I can not tell if it is safe for active write. - not releasing space can be a bug - did you check jira on whamcloud? What version of lustre do you have? Is it ldiskfs or zfs based? zfs version? Alex. From: lustre-discuss mailto:lustre-discuss-boun...@lists.lustre.org>> on behalf of Jason Williams mailto:jas...@jhu.edu>> Sent: Wednesday, January 16, 2019 10:25 AM To: lustr
Re: [lustre-discuss] Migrating files doesn't free space on the OST
Hello Alexander, Thank you for your reply. - We are not using zfs, it's an LDISKFS backing store, so no snapshots. - I have re-run lfs getstripe to make sure the file is indeed moving - I just looked for lfsck but I don't seem to have it. We are running 2.10.4 so I don't know what version that appeared in. - I will try to have a look into the jobstats and see what I can find, but I made sure the files I moved were not in use when I moved them. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: Alexander I Kulyavtsev Sent: Thursday, January 17, 2019 12:56 PM To: Jason Williams; lustre-discuss@lists.lustre.org Subject: Re: Migrating files doesn't free space on the OST - you can re-run command to find files residing on ost to see if files are new or old. - zfs may have snapshots if you ever did snapshots; it takes space. - removing data or snapshots has some lag to release the blocks (tens of minutes) but I guess that is completed by now. - there are can be orphan objects on OST if you had crashes. On older lustre versions if the ost was emptied out you can mount underlying fs as ext4 or zfs; set mount to readonly and browse ost objects - you may see if there are some orphan objects left. On newer lustre releases you probably can run lfsck (lustre scanner). - to find what hosts / jobs currently writing to lustre you may enable lustre jobstats; clear counters and parse stats files in /proc . There was xltop tool on github for older versions of lustre not having implemented jobstats but it was not updated for a while. - depending on lustre version you have the implementation of lfs migrate is different. The older version copied file with other name to other ost, renamed files and removed old file. If migration done on file open for write by application the data will not be released until file closed (and data in new file are wrong). Recent implementation of migrate does swap of the file objects with file layout lock taken. I can not tell if it is safe for active write. - not releasing space can be a bug - did you check jira on whamcloud? What version of lustre do you have? Is it ldiskfs or zfs based? zfs version? Alex. From: lustre-discuss on behalf of Jason Williams Sent: Wednesday, January 16, 2019 10:25 AM To: lustre-discuss@lists.lustre.org Subject: [lustre-discuss] Migrating files doesn't free space on the OST I am trying to migrate files I know are not in use off of the full OST that I have using lfs migrate. I have verified up and down that the files I am moving are on that OST and that after the migrate lfs getstripe indeed shows they are no longer on that OST since it's disabled in the MDS. The problem is, the used space on the OST is not going down. I see one of at least two issues: - the OST is just not freeing the space for some reason or another ( I don't know) - Or someone is writing to existing files just as fast as I am clearing the data (possible, but kind of hard to find) Is there possibly something else I am missing? Also, does anyone know a good way to see if some client is writing to that OST and determine who it is if it's more probable that that is what is going on? -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] Migrating files doesn't free space on the OST
I am trying to migrate files I know are not in use off of the full OST that I have using lfs migrate. I have verified up and down that the files I am moving are on that OST and that after the migrate lfs getstripe indeed shows they are no longer on that OST since it's disabled in the MDS. The problem is, the used space on the OST is not going down. I see one of at least two issues: - the OST is just not freeing the space for some reason or another ( I don't know) - Or someone is writing to existing files just as fast as I am clearing the data (possible, but kind of hard to find) Is there possibly something else I am missing? Also, does anyone know a good way to see if some client is writing to that OST and determine who it is if it's more probable that that is what is going on? -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] lfs_migrate issue
Hello Ahmed, I'm rather new to the lfs_migrate command as well, but one thing to double check is after you run the migrate, have you done an lfs getstripe for a couple of the files it thinks it migrated to make sure they moved off of the OST? Also, did you properly disable the OST in the MDS to make sure new files were not written to it? Here is the document I was following: http://wiki.lustre.org/Handling_Full_OSTs -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: lustre-discuss on behalf of Ahmed Fahmy Sent: Tuesday, January 15, 2019 6:33:24 AM To: lustre-discuss@lists.lustre.org Cc: Supercomputer Subject: [lustre-discuss] lfs_migrate issue Good day, I have been trying to migrate the data one of the OSTs in my lustre file system before removing the OST. I have used the following command: lfs find --obd lustre-OST0001_UUID /lustre | lfs_migrate -sy I believe the data has been copied correctly. However, when I check the size of the directories that has been participating in the migration process, I notice that the size has increased with a 3 or 4 GB increase, even after removing the OST. I am not sure what is the reason for this issue and how I can return those directories to their original sizes. I am using lustre 2.10.1. Any help would be appreciated. Regards, Ahmed Fahmy Bibliotheca Alexandrina ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Full OST, any way of avoiding it without hanging?
Thank you again Rick. One last question, How safe is lfs_migrate? The man page on the installation says it's UNSAFE for possibly in-use files. The lustre manual doesn't have the same warning and says something about it being a bit more integrated with the MDS. http://doc.lustre.org/lustre_manual.xhtml#dbdoclet.lfs_migrate How safe do you think this would be to run on some files on the OST with it disabled on the MDS and active jobs running on the cluster? I could do this by group, possibly to mitigate concerns of open files, if need be. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: Mohr Jr, Richard Frank (Rick Mohr) Sent: Monday, January 7, 2019 1:07 PM To: Jason Williams Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? > On Jan 7, 2019, at 12:53 PM, Jason Williams wrote: > > As I have gone through the testing, I think you may be right. I think I > disabled the OST in a slightly different way and that caused issues. > > Do you happen to know where I could find out a bit more about what the "lctl > set_param osp..max_create_count=0” command would do? The Lustre manual has a section on removing MDTs/OSTs: http://doc.lustre.org/lustre_manual.xhtml#dbdoclet.deactivating_mdt_ost -- Rick Mohr Senior HPC System Administrator National Institute for Computational Sciences http://www.nics.tennessee.edu ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Full OST, any way of avoiding it without hanging?
As I have gone through the testing, I think you may be right. I think I disabled the OST in a slightly different way and that caused issues. Do you happen to know where I could find out a bit more about what the "lctl set_param osp..max_create_count=0” command would do? -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: Mohr Jr, Richard Frank (Rick Mohr) Sent: Monday, January 7, 2019 12:35 PM To: Jason Williams Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? Jason, The results you described sound like the correct behavior when you deactivate an OST on the MDS. When you run “lctl —device deactivate”, you are essentially telling the MDS to ignore that OST when it assigns stripes to a new file. The OST will still be visible to all clients and the MDS, which allows the clients to keep reading files from that OST and allows you to delete files from the OST. The only down side is that any file that already exists on that OST can still be written to. Deactivating an OST is intended to stop the flow of new data to that OST while you work on removing some of the existing data, but it doesn’t actually make the OST read-only. I think you can get the same effect from Lustre 2.9 (or newer) by using "lctl set_param osp..max_create_count=0”. I suspect that what you originally did was to deactivate the OST using something like "lctl conf_param .osc.active=0”. This will notify all Lustre clients to deactivate the OST, which I believe causes the hangs you were seeing when any client tries to remove or stat a file on that OST. -- Rick Mohr Senior HPC System Administrator National Institute for Computational Sciences http://www.nics.tennessee.edu > On Jan 7, 2019, at 11:56 AM, Jason Williams wrote: > > Sorry for the spam, but here are a few more interesting results: > > If I create a file that stripes only on the full OST, and then disable the > OST, I get the following: > >• I can over write the file within it's original size and it takes up > space on the "disabled" OST. >• I can zero the file. >• I can write more data to the file than it originally had, ie. the > original file before disabling the OST was 1G, I can overwrite the file with > > 1G with the OST disabled. >• If I create a new file asking for that OST with the OST disabled, I > get a different OST > > > # 4 and #2 are the only expected behavior. I'm not sure what the behavior > should be in the case of #1 and #3. > > > -- > Jason Williams > Assistant Director > Systems and Data Center Operations. > Maryland Advanced Research Computing Center (MARCC) > Johns Hopkins University > jas...@jhu.edu > > > From: lustre-discuss on behalf of > Jason Williams > Sent: Monday, January 7, 2019 11:47:09 AM > To: Mohr Jr, Richard Frank (Rick Mohr) > Cc: lustre-discuss@lists.lustre.org > Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without > hanging? > > So I found this: http://wiki.lustre.org/Handling_Full_OSTs which is what I > thought I had followed before but ran into hang issues. I did some quick > testing with this and found that: > > 1. if I deactivate the OST in the MDS, no new files appear to be created on > that OST (expected behavior) and no hangs. > 2. If I first create a file on the OST with it activated, then deactivate the > OST, and OVERWRITE a file what was spanned on that OST, the indexes stay the > same and the file successfully overwrites (the file spanned 4 OSTs, so > perhaps a little more testing with a single OST in the index is necessary) > 3. Deactivating the OST shows it as inactive in the MDS but UP in the Client. > (not expected.) > 4. I am able to delete a file that spans that OST with the OST deactivated, > no hang. > > I think the only thing here that concerns me a bit is #2. > > -- > Jason Williams > Assistant Director > Systems and Data Center Operations. > Maryland Advanced Research Computing Center (MARCC) > Johns Hopkins University > jas...@jhu.edu > > > From: lustre-discuss on behalf of > Jason Williams > Sent: Sunday, January 6, 2019 5:22:16 PM > To: Mohr Jr, Richard Frank (Rick Mohr) > Cc: lustre-discuss@lists.lustre.org > Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without > hanging? > > Hi Rick, > I thought what I had done was disable it on the MDS, but perhaps I was > following the wrong instructions. Do you know where the best instructions for > what you are describing can be found? I would be willing
Re: [lustre-discuss] Full OST, any way of avoiding it without hanging?
Sorry for the spam, but here are a few more interesting results: If I create a file that stripes only on the full OST, and then disable the OST, I get the following: 1. I can over write the file within it's original size and it takes up space on the "disabled" OST. 2. I can zero the file. 3. I can write more data to the file than it originally had, ie. the original file before disabling the OST was 1G, I can overwrite the file with > 1G with the OST disabled. 4. If I create a new file asking for that OST with the OST disabled, I get a different OST # 4 and #2 are the only expected behavior. I'm not sure what the behavior should be in the case of #1 and #3. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: lustre-discuss on behalf of Jason Williams Sent: Monday, January 7, 2019 11:47:09 AM To: Mohr Jr, Richard Frank (Rick Mohr) Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? So I found this: http://wiki.lustre.org/Handling_Full_OSTs which is what I thought I had followed before but ran into hang issues. I did some quick testing with this and found that: 1. if I deactivate the OST in the MDS, no new files appear to be created on that OST (expected behavior) and no hangs. 2. If I first create a file on the OST with it activated, then deactivate the OST, and OVERWRITE a file what was spanned on that OST, the indexes stay the same and the file successfully overwrites (the file spanned 4 OSTs, so perhaps a little more testing with a single OST in the index is necessary) 3. Deactivating the OST shows it as inactive in the MDS but UP in the Client. (not expected.) 4. I am able to delete a file that spans that OST with the OST deactivated, no hang. I think the only thing here that concerns me a bit is #2. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: lustre-discuss on behalf of Jason Williams Sent: Sunday, January 6, 2019 5:22:16 PM To: Mohr Jr, Richard Frank (Rick Mohr) Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? Hi Rick, I thought what I had done was disable it on the MDS, but perhaps I was following the wrong instructions. Do you know where the best instructions for what you are describing can be found? I would be willing to try again. — Sent you tersely from my phone Jason Williams From: Mohr Jr, Richard Frank (Rick Mohr) Sent: Sunday, January 6, 2019 4:56 PM To: Jason Williams Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? > On Jan 5, 2019, at 9:49 PM, Jason Williams wrote: > > I have looked around the internet and found you can disable an OST, but when > I have tried that, any writes (including deletes) to the OST hang the clients > indefinitely. Does anyone know a way to make an OST basically "read-only" > with the exception of deletes so we can work to clear out the OST? What command did you use to disable the OST? There is a way to disable the OST on all the clients, but there is also a way to deactivate it on the MDS. The latter method should prevent the MDS from allocating any new files to the OST, but still allow clients to read and delete files on that OST. -- Rick Mohr Senior HPC System Administrator National Institute for Computational Sciences http://www.nics.tennessee.edu ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Full OST, any way of avoiding it without hanging?
So I found this: http://wiki.lustre.org/Handling_Full_OSTs which is what I thought I had followed before but ran into hang issues. I did some quick testing with this and found that: 1. if I deactivate the OST in the MDS, no new files appear to be created on that OST (expected behavior) and no hangs. 2. If I first create a file on the OST with it activated, then deactivate the OST, and OVERWRITE a file what was spanned on that OST, the indexes stay the same and the file successfully overwrites (the file spanned 4 OSTs, so perhaps a little more testing with a single OST in the index is necessary) 3. Deactivating the OST shows it as inactive in the MDS but UP in the Client. (not expected.) 4. I am able to delete a file that spans that OST with the OST deactivated, no hang. I think the only thing here that concerns me a bit is #2. -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> From: lustre-discuss on behalf of Jason Williams Sent: Sunday, January 6, 2019 5:22:16 PM To: Mohr Jr, Richard Frank (Rick Mohr) Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? Hi Rick, I thought what I had done was disable it on the MDS, but perhaps I was following the wrong instructions. Do you know where the best instructions for what you are describing can be found? I would be willing to try again. — Sent you tersely from my phone Jason Williams From: Mohr Jr, Richard Frank (Rick Mohr) Sent: Sunday, January 6, 2019 4:56 PM To: Jason Williams Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? > On Jan 5, 2019, at 9:49 PM, Jason Williams wrote: > > I have looked around the internet and found you can disable an OST, but when > I have tried that, any writes (including deletes) to the OST hang the clients > indefinitely. Does anyone know a way to make an OST basically "read-only" > with the exception of deletes so we can work to clear out the OST? What command did you use to disable the OST? There is a way to disable the OST on all the clients, but there is also a way to deactivate it on the MDS. The latter method should prevent the MDS from allocating any new files to the OST, but still allow clients to read and delete files on that OST. -- Rick Mohr Senior HPC System Administrator National Institute for Computational Sciences http://www.nics.tennessee.edu ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Full OST, any way of avoiding it without hanging?
Hi Rick, I thought what I had done was disable it on the MDS, but perhaps I was following the wrong instructions. Do you know where the best instructions for what you are describing can be found? I would be willing to try again. — Sent you tersely from my phone Jason Williams From: Mohr Jr, Richard Frank (Rick Mohr) Sent: Sunday, January 6, 2019 4:56 PM To: Jason Williams Cc: lustre-discuss@lists.lustre.org Subject: Re: [lustre-discuss] Full OST, any way of avoiding it without hanging? > On Jan 5, 2019, at 9:49 PM, Jason Williams wrote: > > I have looked around the internet and found you can disable an OST, but when > I have tried that, any writes (including deletes) to the OST hang the clients > indefinitely. Does anyone know a way to make an OST basically "read-only" > with the exception of deletes so we can work to clear out the OST? What command did you use to disable the OST? There is a way to disable the OST on all the clients, but there is also a way to deactivate it on the MDS. The latter method should prevent the MDS from allocating any new files to the OST, but still allow clients to read and delete files on that OST. -- Rick Mohr Senior HPC System Administrator National Institute for Computational Sciences http://www.nics.tennessee.edu ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] Full OST, any way of avoiding it without hanging?
Hello, We have a lustre system (version 2.10.4) that has unfortunately fallen victim to a 100% full OST... Every time we clear some space on it, the system fills it right back up again. I have looked around the internet and found you can disable an OST, but when I have tried that, any writes (including deletes) to the OST hang the clients indefinitely. Does anyone know a way to make an OST basically "read-only" with the exception of deletes so we can work to clear out the OST? Or better yet, a way to "drain" or move files off an OST with a script (keeping in mind it might not be known if the files are in use at the time). Or even a way to tell lustre "Hey don't write any new data here, but reading and removing data is OK." -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] Quota Reporting (for all users and/or gruops)
It is entirely possible that this already exists, but my google-foo is not what it used to be. However, I've searched around the internet and it seems as though it doesn't really exist. There are handfuls of now defunct or un-maintained projects out there, but nothing that seems to report all of the user and/or group quotas. Does anyone know of a good quota reporting tool that can give quota information in the same way as a 'repquota -u' or 'repquota -g' would? -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] Spiking OSS load?
Hello, First off, the Lustre that we run here is one that was installed by Intel, so figuring out the exact version seems to be a table lookup on a table internal to Intel, but I'm told it's probably 2.5-ish... Recently, I finally installed a monitoring system on my OSS/MDS servers. And over the last week or so, the OSS servers have been spiking to a 100+ load average (sometimes much higher.) They are not going unresponsive, from what I can tell, and the processes that are causing it seem to be the ll_ost_io07_XXX processes (where XXX is a number) because they are going into "D" state (io wait state) I recently attended the "The 3rd International Workshop on the Lustre Ecosystem" (GREAT WORKSHOP!!) and via a couple of the talks it got me thinking about tunables. One tunable in particular was the ost.threads_max. That guidance on the lustre.org says (1/128MB * num_cpu) which, on my system, works out to well over the max thread count allowable of 512. So my OSS machines are all set to a threads_max of 512 and indeed on all of the machines, the threads_started is 512. (It's a VERY busy file system) This leads me to the following questions (and possibly more, but let's start with these): 1) Is 512 threads a reasonable setting or should it be lower? 2) Is high load "normal" if the file system is under heavy use? At the time I see a lot of open and attr calls which I thought would load the MDS over the OSS... but my under-the-hood understanding is limited at best. 3) Should I be looking at other tunables? I realize the information provided in this initial email is limited as well, so if you are curious about anything else, please let me know what else might be interesting. Oh and as for the MDS/OSS setup, here's a brief overview too: 2x MDS in failover mode with one MDT 12x OSS in fail over pairs with 12 OST per pair 6 running on each OSS. (72 OST, 6 active on each of 12 OSS for load sharing.) Each OSS pair is hooked to the same set of 2x RAID Array (Dell MD3460) -- Jason Williams Assistant Director Systems and Data Center Operations. Maryland Advanced Research Computing Center (MARCC) Johns Hopkins University jas...@jhu.edu<mailto:jas...@jhu.edu> ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [Lustre-discuss] Kernel Oops on the stock RHEL 4 kernel?
Guy Coates wrote: Jason Williams wrote: Hi I have been playing around with lustre 1.6.5.1 as part of some testing that we are doing for an up and coming cluster. I installed it on 2 test machines, Dell 2950's with 8 GB of ram to be exact, and fired up a test file system. The test file system was very simple: /dev/sdb - ~400GB for the MDT/MGS /dev/sdc - ~4TB for the OST Quick questions; did you mount the filesystem on the OST/MDS machine, or on a separate client? Mounting filesystems on OST/MDS nodes is not supported. Does it work with an OST disk 2TB? support for disks 2TB is patchy depending on your exact disk controller hardware. Cheers, Guy Hi Guy Hmm, yes the file system is mounting on the OST/MDS node.Lustre 1.4 seemed to not really have any issue with that. I wonder why it's not supported. And thanks for the heads up on the 2TB support. Looks like I get to go back to my boss and have a discussion about alternatives -- Jason ___ Lustre-discuss mailing list Lustre-discuss@lists.lustre.org http://lists.lustre.org/mailman/listinfo/lustre-discuss
Re: [Lustre-discuss] Kernel Oops on the stock RHEL 4 kernel?
Brian J. Murrell wrote: On Fri, 2008-10-10 at 08:32 -0400, Jason Williams wrote: Hi Hi. --- [cut here ] - [please bite here ] - Kernel BUG at mballoc:1334 This looks like bug 16101 fixed in 1.6.6. There is a patch in that bug you can apply if you wish or you can wait for 1.6.6. Before you ask though, I don't know when 1.6.6 will be released. b. ___ Lustre-discuss mailing list Lustre-discuss@lists.lustre.org http://lists.lustre.org/mailman/listinfo/lustre-discuss Brian, What about Guy's comments about running the OST on a machine that also has the filesystem mounted via lustre client? Is that still technically unsupported in 1.6.6? -- Jason ___ Lustre-discuss mailing list Lustre-discuss@lists.lustre.org http://lists.lustre.org/mailman/listinfo/lustre-discuss