[Gluster-devel] spurious regression failures again! [bug-1112559.t]
Joseph Fernandes
josferna at redhat.com
Thu Jul 24 06:21:37 UTC 2014
Hi All,
After further investigation we have the root cause for this issue.
The root cause is the way in which a new node is added to the cluster.
Now we have N1(127.1.1.1) and N2(127.1.1.2) as two nodes in the cluster, each having a brick N1:B1 (127.1.1.1 : 49146) and N2:B2 (127.1.1.2 : 49147)
Now lets peer probe N3(127.1.1.3) from N1
1) Friend request is sent from N1 to N3. N3 added N1 in the peerinfo list i.e N1 and its uuid say [UUID1]
2) N3 get the brick infos from N1
3) N3 tries to start the bricks
1) N3 tries to start the brick B1 and find its not a local brick, using the logic MY_UUID == brickinfo->uuid, which is false in this case,
as the UUID of brickinfo->hostname (N1) is [UUID1] (as suggested by the peerinfo list) and MY_UUID is [UUID3]. Hence doesn't start it.
2) N3 tries to start the brick B2. Now the problem lies here. N3 uses glusterd_resolve_brick() to resolve the UUID of B2->hostname(N2).
In glusterd_resolve_brick(), it cannot find N2 in the peerinfo list. Then it checks if N2 is a local loop back address. Since N2(127.1.1.2) starts with
"127" it decides that its a local loop back address. Thus glusterd_resolve_brick() fills brickinfo->uuid with [UUID3]. Now as brickinfo->uuid == MY_UUID is
true, N3 initiates the brick process B2 with -s 127.1.1.2 and *-posix.glusterd-uuid=[UUID3]. This process dies off immediately, But for a short amount of
time it holds on to the --brick-port, say for example 49155
All the above is observed & inferred from glusterd logs from N3 (by adding some extra debug messages)
Now coming back to our test case, i.e firing snapshot create and peer probe together. If N2 has assigned 49155 as the port --brick-port for the snapshot brick, then it finds that 49155 is Already acquired by some other process(i.e faulty brick process N3:B2 (127.1.1.2 : 49155), which as the -s 127.1.1.2 and *-posix.glusterd-uuid=[UUID3]) and hence fails to start the snapshot brick process.
1) The error is spurious, as its all about chance when N2 and N3 use the same port for their brick processes.
2) This issue is possible only in a regression test scenario, As all the nodes are on the same machine, differentiated only by a different loop back address (127.1.1.*).
3) Plus The logic that "127" is a local loop back address is also not wrong as glusterd's are suppose to run on different machines in real usage cases.
Please do share your thoughts on this, And what would be a possible fix.
Regards,
Joe
----- Original Message -----
From: "Joseph Fernandes" <josferna at redhat.com>
To: "Avra Sengupta" <asengupt at redhat.com>, "Gluster Devel" <gluster-devel at gluster.org>
Sent: Tuesday, July 22, 2014 6:42:02 PM
Subject: Re: [Gluster-devel] spurious regression failures again! [bug-1112559.t]
Hi All,
As with further investigation found the following,
1) Was the able to reproduce the issue, without running the complete regression, just by running bug-1112559.t only on slave30(which is been rebooted and a clean gluster setup).
This rules out any involvement of previous failure from other spurious errors like mgmt_v3-locks.t.
2) Added some messages and script (netstat and ps -ef | grep gluster ) execution when the binding to a port fails (in rpc/rpc-transport/socket/src/socket.c) and found the following,
Always the snapshot brick in second node (127.1.1.2) fails to acquire the port (eg : 127.1.1.2 : 49155 )
Netstat output shows:
tcp 0 0 127.1.1.2:49155 0.0.0.0:* LISTEN 3555/glusterfsd
and the process that is holding the port 49155 is
root 3555 1 0 12:38 ? 00:00:00 /usr/local/sbin/glusterfsd -s 127.1.1.2 --volfile-id patchy.127.1.1.2.d-backends-2-patchy_snap_mnt -p /d/backends/3/glusterd/vols/patchy/run/127.1.1.2-d-backends-2-patchy_snap_mnt.pid -S /var/run/ff772f1ff85950660f389b0ed43ba2b7.socket --brick-name /d/backends/2/patchy_snap_mnt -l /usr/local/var/log/glusterfs/bricks/d-backends-2-patchy_snap_mnt.log --xlator-option *-posix.glusterd-uuid=3af134ec-5552-440f-ad24-1811308ca3a8 --brick-port 49155 --xlator-option patchy-server.listen-port=49155
Please note even though it says 127.1.1.2 its shows the glusterd-uuid of the 3 node that was been probed when the snapshot was created "3af134ec-5552-440f-ad24-1811308ca3a8"
To clarify things there, there are already a volume brick in 127.1.1.2
root 3446 1 0 12:38 ? 00:00:00 /usr/local/sbin/glusterfsd -s 127.1.1.2 --volfile-id patchy.127.1.1.2.d-backends-2-patchy_snap_mnt -p /d/backends/2/glusterd/vols/patchy/run/127.1.1.2-d-backends-2-patchy_snap_mnt.pid -S /var/run/e667c69aa7a1481c7bd567b917cd1b05.socket --brick-name /d/backends/2/patchy_snap_mnt -l /usr/local/var/log/glusterfs/bricks/d-backends-2-patchy_snap_mnt.log --xlator-option *-posix.glusterd-uuid=a7f461d0-5ea7-4b25-b6c5-388d8eb1893f --brick-port 49153 --xlator-option patchy-server.listen-port=49153
And the above brick process(3555) is not visible before the snap creation or after the failure of the snap brick start on the 127.1.1.2
This means that this process was spawned and died during the creation of the snapshot and probe of the 3rd node (which happens simultaneously)
In addition to these process, we can see multiple snap brick process for the second brick on second node, which are not seen after the failure to start the snap brick on 127.1.1.2
root 3582 1 0 12:38 ? 00:00:00 /usr/local/sbin/glusterfsd -s 127.1.1.2 --volfile-id /snaps/patchy_snap1/66ac70130be84b5c9695df8252a56a6d.127.1.1.2.var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2 -p /d/backends/2/glusterd/snaps/patchy_snap1/66ac70130be84b5c9695df8252a56a6d/run/127.1.1.2-var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2.pid -S /var/run/668f3d4b1c55477fd5ad1ae381de0447.socket --brick-name /var/run/gluster/snaps/66ac70130be84b5c9695df8252a56a6d/brick2 -l /usr/local/var/log/glusterfs/bricks/var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2.log --xlator-option *-posix.glusterd-uuid=a7f461d0-5ea7-4b25-b6c5-388d8eb1893f --brick-port 49155 --xlator-option 66ac70130be84b5c9695df8252a56a6d-server.listen-port=49155
root 3583 3582 0 12:38 ? 00:00:00 /usr/local/sbin/glusterfsd -s 127.1.1.2 --volfile-id /snaps/patchy_snap1/66ac70130be84b5c9695df8252a56a6d.127.1.1.2.var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2 -p /d/backends/2/glusterd/snaps/patchy_snap1/66ac70130be84b5c9695df8252a56a6d/run/127.1.1.2-var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2.pid -S /var/run/668f3d4b1c55477fd5ad1ae381de0447.socket --brick-name /var/run/gluster/snaps/66ac70130be84b5c9695df8252a56a6d/brick2 -l /usr/local/var/log/glusterfs/bricks/var-run-gluster-snaps-66ac70130be84b5c9695df8252a56a6d-brick2.log --xlator-option *-posix.glusterd-uuid=a7f461d0-5ea7-4b25-b6c5-388d8eb1893f --brick-port 49155 --xlator-option 66ac70130be84b5c9695df8252a56a6d-server.listen-port=49155
This looks like the second node tries to start snap brick
1) with wrong brickinfo and peerinfo (process 3555)
2) Multiple times with the correct brickinfo (process 3582,3583)
3) This issue is not seen when, snapshots are created and peer probe is NOT done simultaneously.
Will continue on the investigation and will keep you posted.
Regards,
Joe
----- Original Message -----
From: "Joseph Fernandes" <josferna at redhat.com>
To: "Avra Sengupta" <asengupt at redhat.com>
Cc: "Pranith Kumar Karampuri" <pkarampu at redhat.com>, "Gluster Devel" <gluster-devel at gluster.org>, "Varun Shastry" <vshastry at redhat.com>, "Justin Clift" <justin at gluster.org>
Sent: Thursday, July 17, 2014 10:58:14 AM
Subject: Re: [Gluster-devel] spurious regression failures again!
Hi Avra,
Just clarifying things here,
1) When testing with the setup provide by Justin, I found the only place where bug-1112559.t failed was after the failure mgmt_v3-locks.t in the previous regression run. The mail attached with the previous mail was just an OBSERVATION and NOT an INFERENCE that failure of mgmt_v3-locks.t was the root cause of bug-1112559.t . I am NOT jumping the gun and making any statement/conclusion here. Its just an OBSERVATION. And thanks for the clarification on why mgmt_v3-locks.t is failing.
2) I agree with you that the cleanup script needs to kill all gluster* processes. And its also true that port range used by gluster for bricks is unique.
But bug-1112559.t fails only because of the unavailability of port, to start the snap brick. Therefore this suggests that there is some process(gluster or non-gluster)
still using the port.
3) And Finally that bug-1112559.t failing individually all the time is not true as when looked into the links which you have provided there are case where there are previous other test case failures, on the same testing machine (slave26). By this I am not pointing out that those failure are the root cause for bug-1112559.t to fail
As stated earlier its a notable OBSERVATION(Keeping in mind point 2 about ports and cleanup)
I have run nearly 30 runs on slave30 and only one time bug-1112559.t failed (As stated in point 1). I am continuing to run more runs. The only problem is the occurrence of bug-1112559.t failure is spurious and there is no deterministic way of reproducing it.
Will keep all posted about the results.
Regards,
Joe
----- Original Message -----
From: "Avra Sengupta" <asengupt at redhat.com>
To: "Joseph Fernandes" <josferna at redhat.com>, "Pranith Kumar Karampuri" <pkarampu at redhat.com>
Cc: "Gluster Devel" <gluster-devel at gluster.org>, "Varun Shastry" <vshastry at redhat.com>, "Justin Clift" <justin at gluster.org>
Sent: Wednesday, July 16, 2014 1:03:21 PM
Subject: Re: [Gluster-devel] spurious regression failures again!
Joseph,
I am not sure I understand how this is affecting the spurious failure of
bug-1112559.t. As per the mail you have attached, and according to your
analysis, bug-1112559.t fails because a cleanup hasn't happened
properly after a previous test-case failed and in your case there was a
crash as well.
Now out of all the times bug-1112559.t has failed, most of the time it's
the only test case failing and there isn't any crash. Below are the
regression runs that pranith had sent for the same.
http://build.gluster.org/job/rackspace-regression-2GB/541/consoleFull
http://build.gluster.org/job/rackspace-regression-2GB-triggered/173/consoleFull
http://build.gluster.org/job/rackspace-regression-2GB-triggered/172/consoleFull
http://build.gluster.org/job/rackspace-regression-2GB/543/console
In all of the above bug-1112559.t is the only test case that fails and
there is no crash.
So what I fail to understand here is, if this particular testcase fails
independently as well as with other testcases, then how can we conclude
that any other testcase failing is somehow not doing a cleanup properly
and that is the reason for bug-1112559.t failing.
mgmt_v3-locks.t fails because glusterd takes more time to register a
node going down, and hence the peer status doesn't return what the
testcase expects it to. It's a race. The testcase ends with a cleanup
routine like every other testcase, that kills all gluster and glusterfsd
processes, which might be using any brick ports. So could you please
explain how or which process still uses the brick ports that the snap
bricks are trying to use leading to the failure of bug-1112559.t.
Regards,
Avra
On 07/15/2014 09:57 PM, Joseph Fernandes wrote:
> Just pointing out ,
>
> 2) tests/basic/mgmt_v3-locks.t - Author: Avra
> http://build.gluster.org/job/rackspace-regression-2GB-triggered/375/consoleFull
>
> This is the similar kind of error I saw in my testing of spurious failure tests/bugs/bug-1112559.t
>
> Please refer the attached mail.
>
> Regards,
> Joe
>
>
>
> ----- Original Message -----
> From: "Pranith Kumar Karampuri" <pkarampu at redhat.com>
> To: "Joseph Fernandes" <josferna at redhat.com>
> Cc: "Gluster Devel" <gluster-devel at gluster.org>, "Varun Shastry" <vshastry at redhat.com>
> Sent: Tuesday, July 15, 2014 9:34:26 PM
> Subject: Re: [Gluster-devel] spurious regression failures again!
>
>
> On 07/15/2014 09:24 PM, Joseph Fernandes wrote:
>> Hi Pranith,
>>
>> Could you please share the link of the console output of the failures.
> Added them inline. Thanks for reminding :-)
>
> Pranith
>> Regards,
>> Joe
>>
>> ----- Original Message -----
>> From: "Pranith Kumar Karampuri" <pkarampu at redhat.com>
>> To: "Gluster Devel" <gluster-devel at gluster.org>, "Varun Shastry" <vshastry at redhat.com>
>> Sent: Tuesday, July 15, 2014 8:52:44 PM
>> Subject: [Gluster-devel] spurious regression failures again!
>>
>> hi,
>> We have 4 tests failing once in a while causing problems:
>> 1) tests/bugs/bug-1087198.t - Author: Varun
> http://build.gluster.org/job/rackspace-regression-2GB-triggered/379/consoleFull
>> 2) tests/basic/mgmt_v3-locks.t - Author: Avra
> http://build.gluster.org/job/rackspace-regression-2GB-triggered/375/consoleFull
>> 3) tests/basic/fops-sanity.t - Author: Pranith
> http://build.gluster.org/job/rackspace-regression-2GB-triggered/383/consoleFull
>> Please take a look at them and post updates.
>>
>> Pranith
>> _______________________________________________
>> Gluster-devel mailing list
>> Gluster-devel at gluster.org
>> http://supercolony.gluster.org/mailman/listinfo/gluster-devel
More information about the Gluster-devel
mailing list