Hi, the grant_shrink bug was fixed in 2.15.0 according to this JIRA: https://jira.whamcloud.com/browse/LU-14124
On Fri, Sep 30, 2022 at 3:59 AM Tung-Han Hsieh < [email protected]> wrote: > Dear Peter, > > Thank you very much for your prompt reply. > > Actually we just encountered OST file server freezed a few hours ago, > which installed Lustre-2.15.0 + zfs-2.0.7. Unfortunately all the kernel > and event logs were lost. So we don't know what's going on. > > This system was upgraded to Lustre-2.15.0 + zfs-2.0.7 two days ago. > Before that it runs Lustre-2.12.6 + zfs-0.7.13 smoothly for two years. > So we suspected that this system crash was due to the Lustre upgrade. > To take some actions, we just upgraded the system to Lustre-2.15.1 + > zfs-2.0.7, and hope that it could run without problem. > > By the way, we did see some minor error messages in dmesg when mounting > MDT and OST in the MDT and OST servers, respectively, with Lustre-2.15.0 > and 2.15.1: > > [ 1049.086838] systemd-udevd[6131]: Process '/usr/sbin/lctl set_param > 'osc.*.grant_shrink=1'' failed with exit code 2. > > I guess this command should be run in the clients, but not valid in MDT > and OST. Our /etc/udev/rules.d/99-lustre.rules reads: > > ======================================================================== > KERNEL=="obd", MODE="0666" > > # set sysfs values on client > SUBSYSTEM=="lustre", ACTION=="change", ENV{PARAM}=="?*", > RUN+="/usr/sbin/lctl set_param '$env{PARAM}=$env{SETTING}'" > # setup linux routes for mr on lustre load > SUBSYSTEM=="module", ACTION=="add", DEVPATH=="/module/lustre", > RUN+="/usr/sbin/lnetctl setup-mrrouting" > ======================================================================== > > and /etc/udev/rules.d/99-lustre-server.rules reads: > > ======================================================================== > # Ensure block devices re-added to the system allow for large writes > (LU-9551) > # Currently applies only to ldiskfs-formatted devices > ACTION!="add|change", GOTO="l_tunedisk_end" > SUBSYSTEM!="block", GOTO="l_tunedisk_end" > ENV{ID_FS_TYPE}!="ext4", GOTO="l_tunedisk_end" > > ENV{ID_FS_LABEL}=="MGS|*-MDT*|*-OST*", RUN+="/usr/sbin/l_tunedisk /dev/%k" > > LABEL="l_tunedisk_end" > ======================================================================== > > Further, when running Lustre-2.12.6 + zfs-0.7.3, we have to manually set > > lctl set_para osc.*.grant_shrink=0 > > in order to maintain reasonable performance when transferring large files. > Is it necessary to do that for Lustre-2.15.X + zfs-2.0.7 ? > > > Cheers, > > T.H.Hsieh > > > On Fri, Sep 30, 2022 at 07:18:53AM +0000, Peter Jones wrote: > > There is a general problem with the Lustre.org wiki ATM. Hopefully > service will be restored shortly. In the meantime is there a specific piece > of information that you wanted to know? The list of fixed issues is also > available in JIRA for example - > https://jira.whamcloud.com/secure/ReleaseNote.jspa?projectId=10000&version=15891 > > > > ???On 2022-09-30, 12:09 AM, "lustre-discuss on behalf of Tung-Han Hsieh" > <[email protected] on behalf of > [email protected]> wrote: > > > > Dear All, > > > > Could anyone point out where to find the Lustre-2.15.1 change log ? > > > > This URL found in Lustre website is invalid: > > > > https://wiki.lustre.org/Lustre_2.15.1_Changelog > > > > Thank you very much. > > > > > > T.H.Hsieh > > _______________________________________________ > > lustre-discuss mailing list > > [email protected] > > http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org > > > _______________________________________________ > lustre-discuss mailing list > [email protected] > http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org >
_______________________________________________ lustre-discuss mailing list [email protected] http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
