[Gluster-users] [External] Re: Self Heal Confusion

[Brett Holcomb]

I wrote a script to search the output of gluster volume heal projects 
info, picks the brick I gave it and then deletes any of the files  
listed that actually exist in .glusterfs/dir1/dir2.  I did this on the 
first host which had 85 pending and that cleared them up so I'll do it 
via ssh on the other two servers.

Hopefully that will clear it up and glusterfs will be happy again.

Thanks everyone for the help.


On 12/31/18 4:39 AM, Davide Obbi wrote:
> cluster.quorum-type auto
> cluster.quorum-count (null)
> cluster.server-quorum-type off
> cluster.server-quorum-ratio 0
> cluster.quorum-reads                    no
>
> Where exacty do I remove the gfid entries from - the .glusterfs
> directory? --> yes can't remember exactly where but try to do a find 
> in the brick paths with the gfid  it should return something
>
> Where do I put the cluster.heal-timeout option - which file? --> 
> gluster volume set volumename option value
>
> On Mon, Dec 31, 2018 at 10:34 AM Brett Holcomb <biholcomb at l1049h.com 
> <mailto:biholcomb at l1049h.com>> wrote:
>
>     That is probably the case as a lot of files were deleted some time
>     ago.
>
>     I'm on version 5.2 but was on 3.12 until about a week ago.
>
>     Here is the quorum info.  I'm running a distributed replicated
>     volumes
>     in 2 x 3 = 6
>
>     cluster.quorum-type auto
>     cluster.quorum-count (null)
>     cluster.server-quorum-type off
>     cluster.server-quorum-ratio 0
>     cluster.quorum-reads                    no
>
>     Where exacty do I remove the gfid entries from - the .glusterfs
>     directory?  Do I just delete all the directories can files under this
>     directory?
>
>     Where do I put the cluster.heal-timeout option - which file?
>
>     I think you've hit on the cause of the issue.  Thinking back we've
>     had
>     some extended power outages and due to a misconfiguration in the swap
>     file device name a couple of the nodes did not come up and I didn't
>     catch it for a while so maybe the deletes occured then.
>
>     Thank you.
>
>     On 12/31/18 2:58 AM, Davide Obbi wrote:
>     > if the long GFID does not correspond to any file it could mean the
>     > file has been deleted by the client mounting the volume. I think
>     this
>     > is caused when the delete was issued and the number of active
>     bricks
>     > were not reaching quorum majority or a second brick was taken down
>     > while another was down or did not finish the selfheal, the
>     latter more
>     > likely.
>     > It would be interesting to see:
>     > - what version of glusterfs you running, it happened to me with 3.12
>     > - volume quorum rules: "gluster volume get vol all | grep quorum"
>     >
>     > To clean it up if i remember correctly it should be possible to
>     delete
>     > the gfid entries from the brick mounts on the glusterfs server
>     nodes
>     > reporting the files to heal.
>     >
>     > As a side note you might want to consider changing the selfheal
>     > timeout to more agressive schedule in cluster.heal-timeout option
>     _______________________________________________
>     Gluster-users mailing list
>     Gluster-users at gluster.org <mailto:Gluster-users at gluster.org>
>     https://lists.gluster.org/mailman/listinfo/gluster-users
>
>
>
> -- 
> Davide Obbi
> System Administrator
>
> Booking.com B.V.
> Vijzelstraat 66-80 Amsterdam 1017HL Netherlands
> Direct +31207031558
> Booking.com <https://www.booking.com/>
> Empowering people to experience the world since 1996
> 43 languages, 214+ offices worldwide, 141,000+ global destinations, 29 
> million reported listings
> Subsidiary of Booking Holdings Inc. (NASDAQ: BKNG)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20190104/313f9b43/attachment.html>