[lustre-discuss] lustre 2.10.* and zfs record size

Andreas Dilger adilger at whamcloud.com
Fri Sep 28 01:21:42 PDT 2018


It is best to leave the MDT recordsize at the default. For most files it will not matter (too small), but for Data on MDT you also don't want a large recordsize.

Cheers, Andreas

On Sep 28, 2018, at 09:17, Robert Redl <robert.redl at lmu.de<mailto:robert.redl at lmu.de>> wrote:


Dear All,

I have a related question. The documentation says nothing about the recommended recordsize for MDTs with ZFS backend. Are there any recommendation? Or is data on MDTs stored in a way that the recordsize does not matter?

Thanks,
Robert

On 09/28/2018 12:54 AM, Andreas Dilger wrote:
Firstly, we don't test anything larger than 1MB.

Secondly, the best recordsize is up to the apication IO pattern. If it is streaming writes, it might be OK. If it is random write then probably not.

Cheers, Andreas

On Sep 27, 2018, at 19:04, Riccardo Veraldi <Riccardo.Veraldi at cnaf.infn.it<mailto:Riccardo.Veraldi at cnaf.infn.it>> wrote:


Hello,

I have always been using ZFS record size to 1MB with Lustre on top of ZFS.  Anyway it is possible to set it to more than 1MB. May this have any benefit in performance and is it recommended to set for example:

echo 4194304 > /sys/module/zfs/parameters/zfs_max_recordsize

zfs set recordsize=4m zpool




Thank you.

Rick


_______________________________________________
lustre-discuss mailing list
lustre-discuss at lists.lustre.org<mailto:lustre-discuss at lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org



_______________________________________________
lustre-discuss mailing list
lustre-discuss at lists.lustre.org<mailto:lustre-discuss at lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org


_______________________________________________
lustre-discuss mailing list
lustre-discuss at lists.lustre.org<mailto:lustre-discuss at lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20180928/3a29db8b/attachment.html>


More information about the lustre-discuss mailing list