[Gluster-users] df shows wrong mount size, after adding bricks to volume

[Petr Certik]

Thanks!

One more question -- I don't really mind having the wrong size
reported by df, but I'm worried whether it is safe to use the volume.
Will it be okay if I write to it? For example, once the actual used
size is greater than the reported total size of the volume, should I
expect problems? And is it safe to upgrade glusterfs when the volume
is in this state?

Cheers,
Petr

On Fri, May 29, 2020 at 11:37 AM Sanju Rakonde <srakonde at redhat.com> wrote:
>
> Nope, for now. I will update you if we figure out any other workaround.
>
> Thanks for your help!
>
> On Fri, May 29, 2020 at 2:50 PM Petr Certik <petr at certik.cz> wrote:
>>
>> I'm afraid I don't have the resources to try and reproduce from the
>> beginning. Is there anything else I can do to get you more
>> information?
>>
>>
>> On Fri, May 29, 2020 at 11:08 AM Sanju Rakonde <srakonde at redhat.com> wrote:
>> >
>> > The issue is not with glusterd restart. We need to reproduce from beginning and add-bricks to check df -h values.
>> >
>> > I suggest not to try on the production environment. if you have any other machines, please let me know.
>> >
>> > On Fri, May 29, 2020 at 1:37 PM Petr Certik <petr at certik.cz> wrote:
>> >>
>> >> If you mean the issue during node restart, then yes, I think I could
>> >> reproduce that with a custom build. It's a production system, though,
>> >> so I'll need to be extremely careful.
>> >>
>> >> We're using debian glusterfs-server 7.3-1 amd64, can you provide a
>> >> custom glusterd binary based off of that version?
>> >>
>> >> Cheers,
>> >> Petr
>> >>
>> >> On Fri, May 29, 2020 at 9:09 AM Sanju Rakonde <srakonde at redhat.com> wrote:
>> >> >
>> >> > Surprising! Will you be able to reproduce the issue and share the logs if I provide a custom build with more logs?
>> >> >
>> >> > On Thu, May 28, 2020 at 1:35 PM Petr Certik <petr at certik.cz> wrote:
>> >> >>
>> >> >> Thanks for your help! Much appreciated.
>> >> >>
>> >> >> The fsid is the same for all bricks:
>> >> >>
>> >> >> imagegluster1:
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data2-brick:brick-fsid=2065
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data-brick:brick-fsid=2065
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data-brick:brick-fsid=0
>> >> >>
>> >> >> imagegluster2:
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data2-brick:brick-fsid=2065
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data-brick:brick-fsid=2065
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data-brick:brick-fsid=0
>> >> >>
>> >> >> imagegluster3:
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster1:-data-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data2-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster2:-data-brick:brick-fsid=0
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data2-brick:brick-fsid=2065
>> >> >> /var/lib/glusterd/vols/gv0/bricks/imagegluster3:-data-brick:brick-fsid=2065
>> >> >>
>> >> >>
>> >> >> I already did try restarting the glusterd nodes with no effect, but
>> >> >> that was before the upgrades of client versions.
>> >> >>
>> >> >> Running the "volume set" command did not seem to work either, the
>> >> >> shared-brick-counts are still the same (2).
>> >> >>
>> >> >> However, when restarting a node, I do get an error and a few warnings
>> >> >> in the log: https://pastebin.com/tqq1FCwZ
>> >> >>
>> >> >>
>> >> >>
>> >> >> On Wed, May 27, 2020 at 3:14 PM Sanju Rakonde <srakonde at redhat.com> wrote:
>> >> >> >
>> >> >> > The shared-brick-count value indicates the number of bricks sharing a file-system. In your case, it should be one, as all the bricks are from different mount points. Can you please share the values of brick-fsid?
>> >> >> >
>> >> >> > grep "brick-fsid" /var/lib/glusterd/vols/<volname>/bricks/
>> >> >> >
>> >> >> > I tried reproducing this issue in fedora vm's but couldn't hit this. we are seeing this issue on and off but are unable to reproduce in-house. If you see any error messages in glusterd.log please share the log too.
>> >> >> >
>> >> >> > Work-around to come out from this situation:
>> >> >> > 1. Restarting the glusterd service on all nodes:
>> >> >> > # systemctl restart glusterd
>> >> >> >
>> >> >> > 2. Run set volume command to update vol file:
>> >> >> > # gluster v set <VOLNAME> min-free-disk 11%
>> >> >> >
>> >> >> > On Wed, May 27, 2020 at 5:24 PM Petr Certik <petr at certik.cz> wrote:
>> >> >> >>
>> >> >> >> As far as I remember, there was no version update on the server. It
>> >> >> >> was definitely installed as version 7.
>> >> >> >>
>> >> >> >> Shared bricks:
>> >> >> >>
>> >> >> >> Server 1:
>> >> >> >>
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data2-brick.vol:
>> >> >> >> option shared-brick-count 2
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data-brick.vol:    option
>> >> >> >> shared-brick-count 2
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >>
>> >> >> >> Server 2:
>> >> >> >>
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data2-brick.vol:
>> >> >> >> option shared-brick-count 2
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data-brick.vol:    option
>> >> >> >> shared-brick-count 2
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >>
>> >> >> >> Server 3:
>> >> >> >>
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster1.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data2-brick.vol:
>> >> >> >> option shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster2.data-brick.vol:    option
>> >> >> >> shared-brick-count 0
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data2-brick.vol:
>> >> >> >> option shared-brick-count 2
>> >> >> >> /var/lib/glusterd/vols/gv0/gv0.imagegluster3.data-brick.vol:    option
>> >> >> >> shared-brick-count 2
>> >> >> >>
>> >> >> >> On Wed, May 27, 2020 at 1:36 PM Sanju Rakonde <srakonde at redhat.com> wrote:
>> >> >> >> >
>> >> >> >> > Hi Petr,
>> >> >> >> >
>> >> >> >> > what was the server version before upgrading to 7.2?
>> >> >> >> >
>> >> >> >> > Can you please share the shared-brick-count values from brick volfiles from all the nodes?
>> >> >> >> > grep shared-brick-count /var/lib/glusterd/vols/<volume_name>/*
>> >> >> >> >
>> >> >> >> > On Wed, May 27, 2020 at 2:31 PM Petr Certik <petr at certik.cz> wrote:
>> >> >> >> >>
>> >> >> >> >> Hi everyone,
>> >> >> >> >>
>> >> >> >> >> we've been running a replicated volume for a while, with three ~1 TB
>> >> >> >> >> bricks. Recently we've added three more same-sized bricks, making it a
>> >> >> >> >> 2 x 3 distributed replicated volume. However, even after rebalance,
>> >> >> >> >> the `df` command on a client shows the correct used/size percentage,
>> >> >> >> >> but wrong absolute sizes. The size still shows up as ~1 TB while in
>> >> >> >> >> reality it should be around 2 TB, and both "used" and "available"
>> >> >> >> >> reported sizes are about half of what they should be. The clients were
>> >> >> >> >> an old version (5.5), but even after upgrade to 7.2 and remount, the
>> >> >> >> >> reported sizes are still wrong. There are no heal entries. What can I
>> >> >> >> >> do to fix this?
>> >> >> >> >>
>> >> >> >> >> OS: debian buster everywhere
>> >> >> >> >> Server version: 7.3-1, opversion: 70200
>> >> >> >> >> Client versions: 5.5-3, 7.6-1, opversions: 50400, 70200
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >> root at imagegluster1:~# gluster volume info gv0
>> >> >> >> >> Volume Name: gv0
>> >> >> >> >> Type: Distributed-Replicate
>> >> >> >> >> Volume ID: 5505d350-9b61-4056-9054-de9dfb58eab7
>> >> >> >> >> Status: Started
>> >> >> >> >> Snapshot Count: 0
>> >> >> >> >> Number of Bricks: 2 x 3 = 6
>> >> >> >> >> Transport-type: tcp
>> >> >> >> >> Bricks:
>> >> >> >> >> Brick1: imagegluster1:/data/brick
>> >> >> >> >> Brick2: imagegluster2:/data/brick
>> >> >> >> >> Brick3: imagegluster3:/data/brick
>> >> >> >> >> Brick4: imagegluster1:/data2/brick
>> >> >> >> >> Brick5: imagegluster2:/data2/brick
>> >> >> >> >> Brick6: imagegluster3:/data2/brick
>> >> >> >> >> Options Reconfigured:
>> >> >> >> >> features.cache-invalidation: on
>> >> >> >> >> transport.address-family: inet
>> >> >> >> >> storage.fips-mode-rchecksum: on
>> >> >> >> >> nfs.disable: on
>> >> >> >> >> performance.client-io-threads: off
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >> root at imagegluster1:~# df -h
>> >> >> >> >> Filesystem      Size  Used Avail Use% Mounted on
>> >> >> >> >> ...
>> >> >> >> >> /dev/sdb1       894G  470G  425G  53% /data2
>> >> >> >> >> /dev/sdc1       894G  469G  426G  53% /data
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >> root at any-of-the-clients:~# df -h
>> >> >> >> >> Filesystem         Size  Used Avail Use% Mounted on
>> >> >> >> >> ...
>> >> >> >> >> imagegluster:/gv0  894G  478G  416G  54% /mnt/gluster
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >> Let me know if there's any other info I can provide about our setup.
>> >> >> >> >>
>> >> >> >> >> Cheers,
>> >> >> >> >> Petr Certik
>> >> >> >> >> ________
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >>
>> >> >> >> >> Community Meeting Calendar:
>> >> >> >> >>
>> >> >> >> >> Schedule -
>> >> >> >> >> Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
>> >> >> >> >> Bridge: https://bluejeans.com/441850968
>> >> >> >> >>
>> >> >> >> >> Gluster-users mailing list
>> >> >> >> >> Gluster-users at gluster.org
>> >> >> >> >> https://lists.gluster.org/mailman/listinfo/gluster-users
>> >> >> >> >>
>> >> >> >> >
>> >> >> >> >
>> >> >> >> > --
>> >> >> >> > Thanks,
>> >> >> >> > Sanju
>> >> >> >>
>> >> >> >
>> >> >> >
>> >> >> > --
>> >> >> > Thanks,
>> >> >> > Sanju
>> >> >>
>> >> >
>> >> >
>> >> > --
>> >> > Thanks,
>> >> > Sanju
>> >>
>> >
>> >
>> > --
>> > Thanks,
>> > Sanju
>>
>
>
> --
> Thanks,
> Sanju