[Gluster-users] error "Not able to add to index" in brick logs

[Stefano Bagnara]

On 22 November 2017 at 22:34, Paul Robert Marino <prmarino1 at gmail.com> wrote:
> Yes indeed it is probably what's going on. what filesystem are you using and what are the mount options?

It's an ext4 with "default" mount options.

# tune2fs -l /dev/sda3 | grep features
Filesystem features:      has_journal ext_attr resize_inode dir_index
filetype needs_recovery extent flex_bg sparse_super large_file
huge_file uninit_bg dir_nlink extra_isize

It is a mirror setup of the "source" cluster that host the same data
on gluster 3.6, so it "should" be something related to a different way
gluster 3.12 works.

May it be this is somehow related to halo and copying millions of
small files so that the "queue" for the halo "geo replica" builds up
to more than 65000 files and this hit a filesystem limit?
I'm disabling halo and starting from scratch..

Stefano

>   Original Message
> From: lists at bago.org
> Sent: November 22, 2017 4:26 PM
> To: gluster-users at gluster.org
> Subject: [Gluster-users] error "Not able to add to index" in brick logs
>
> in my /var/log/gluster/bricks/mybrick-path.log I get thousands of those errors:
> ------
> [2017-11-22 21:06:23.768354] E [MSGID: 138003]
> [index.c:624:index_link_to_base] 0-sharedvol-index:
> /home/sharedvol/.glusterfs/indices/xattrop/0b852dad-b332-4bfe-a38b-976729ee46a2:
> Not able to add to index [Troppi collegamenti]
> The message "E [MSGID: 138003] [index.c:624:index_link_to_base]
> 0-sharedvol-index:
> /home/sharedvol/.glusterfs/indices/xattrop/0b852dad-b332-4bfe-a38b-976729ee46a2:
> Not able to add to index [Troppi collegamenti]" repeated 29 times
> between [2017-11-22 21:06:23.768354] and [2017-11-22 21:06:23.776437]
> --------
>
> "Troppi collegamenti" is the italian for "too many links", I guess
> this is an underlying OS filesystem error preventing gluster from
> doing something.
>
> What does this mean?
>
> find /home/sharedvol/.glusterfs/indices/xattrop |wc -l
> 65001
>
> My brick is on ext4. Gluster is 3.12.1-2 on Centos7, an "old style"
> replica 2 on 2 bricks. I'm pushing millions of files from another
> gluster cluster (that is on 3.6 and i'm unable to upgrade, so I
> decided to create a new cluster and move data).
>
> Stefano