[Bugs] [Bug 1717757] BItrot: Segmentation Fault if bitrot stub do signature

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1717757



--- Comment #4 from david.spisla at iternity.com ---
@Amar

I wrote a patch with debug logs and I will observe the bricks now. During this
time I have some questions concerning your patch suggestion:

1. According to crash report from the brick locks, there was a failure in

[2019-06-05 09:05:05.137156] I [MSGID: 139001]
[posix-acl.c:263:posix_acl_log_permit_denied] 0-archive1-access-control:
client: CTX_ID:fcab5e67-b9d9-4b72-8c15-f29de2084af3-GRAPH_ID:0-PID:189
16-HOST:fs-detlefh-c1-n2-PC_NAME:archive1-client-0-RECON_NO:-0, gfid:
494b42ad-7e40-4e27-8878-99387a80b5dc, req(uid:2000,gid:2000,perm:3,ngrps:1),
ctx(uid:0,gid:0,in-groups:0,perm:755,update
d-fop:LOOKUP, acl:-) [Permission denied]

just before the crash. What can be the reason for this?

2. If this LOOKUP for acls fails, is it problematic to do a setxattr with loc?
If we skip setting xattr when fd is NULL the file on that brick won't have the
necessary xattr like trusted.worm_file and other. See an example directly after
the crash:
# file: gluster/brick3/glusterbrick/test/data/BC/storage.log
trusted.gfid=0sag3y6RuoTgqAw//fx3ZB1Q==
trusted.gfid2path.273f2255a25b2961="bd910b86-d51a-4006-a2c4-515ef5f1777a/storage.log"
trusted.pgfid.bd910b86-d51a-4006-a2c4-515ef5f1777a=0sAAAAAQ==

On the healthy brick I got:
# file: gluster/brick3/glusterbrick/test/data/BC/storage.log
trusted.afr.dirty=0sAAAAAAAAAAAAAAAA
trusted.afr.test-client-0=0sAAAABAAAAAMAAAAA
trusted.bit-rot.version=0sAgAAAAAAAABc+P64AAEhGQ==
trusted.gfid=0sag3y6RuoTgqAw//fx3ZB1Q==
trusted.gfid2path.273f2255a25b2961="bd910b86-d51a-4006-a2c4-515ef5f1777a/storage.log"
trusted.glusterfs.mdata=0sAQAAAAAAAAAAAAAAAFz5AJEAAAAAMqdgMwAAAABcRwJEAAAAAAAAAAAAAAAAXPkAkQAAAAAAAAAA
trusted.pgfid.bd910b86-d51a-4006-a2c4-515ef5f1777a=0sAAAAAQ==
trusted.start_time="1559822481"
trusted.worm_file=0sMQA=

After restarting the faulty brick a heal was triggered and afterwards the file
on the faulty brick is heal.It should be ensured that the broken file gets all
necessary xattr. What is the better way? Triggering a setxattr with loc in
worm_create_cbk or do a heal afterwards?

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the Docs Contact for the bug.