[Gluster-users] rebalance fix layout necessary

Thu Apr 6 09:26:13 UTC 2017

Hi,

I was able to add bricks to the volume successfully.
Client was reading, writing and listing data from mount point.
But after adding bricks I had issues in folder listing (not listing all
folders or returning empty folder list) and write was interrupted.
remounting volume has solved the issue and now working fine.

I was under the impression that running rebalance would cause folder
listing issue but now adding brick itself created a problem.
It's irrelevant whether client busy or idle need to remount to solve the
issue.

Also, i would like to know using brick in a volume without fix-layout cause
folder listing slowness.

Below a snippet of log from client when this happened. let me know if you
any more additional info.

Client and Servers are 3.10.1, volume mounted thru fuse.

Machine busy downloading & uploading

[2017-04-05 13:39:33.487176] I [MSGID: 114021] [client.c:2361:notify]
0-gfs-vol-client-1107: current graph is no longer active, destroying
rpc_client
[2017-04-05 13:39:33.487196] I [MSGID: 114021] [client.c:2361:notify]
0-gfs-vol-client-1108: current graph is no longer active, destroying
rpc_client
[2017-04-05 13:39:33.487201] I [MSGID: 114018]
[client.c:2276:client_rpc_notify] 0-gfs-vol-client-1107: disconnected from
gfs-vol-client-1107. Client process will keep trying to connect to glusterd
until brick's port is available
[2017-04-05 13:39:33.487212] I [MSGID: 114021] [client.c:2361:notify]
0-gfs-vol-client-1109: current graph is no longer active, destroying
rpc_client
[2017-04-05 13:39:33.487217] I [MSGID: 114018]
[client.c:2276:client_rpc_notify] 0-gfs-vol-client-1108: disconnected from
gfs-vol-client-1108. Client process will keep trying to connect to glusterd
until brick's port is available
[2017-04-05 13:39:33.487232] I [MSGID: 114018]
[client.c:2276:client_rpc_notify] 0-gfs-vol-client-1109: disconnected from
gfs-vol-client-1109. Client process will keep trying to connect to glusterd
until brick's port is available

Idle system

2017-04-05 13:40:07.692336] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1065:
Server lk version = 1
[2017-04-05 13:40:07.692383] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-995:
Server lk version = 1
[2017-04-05 13:40:07.692430] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-965:
Server lk version = 1
[2017-04-05 13:40:07.692485] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1075:
Server lk version = 1
[2017-04-05 13:40:07.692532] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1025:
Server lk version = 1
[2017-04-05 13:40:07.692569] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1055:
Server lk version = 1
[2017-04-05 13:40:07.692620] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-955:
Server lk version = 1
[2017-04-05 13:40:07.692681] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1035:
Server lk version = 1
[2017-04-05 13:40:07.692870] I [MSGID: 114035]
[client-handshake.c:202:client_set_lk_version_cbk] 2-gfs-vol-client-1045:
Server lk version = 1

Regards,
Amudhan

On Tue, Apr 4, 2017 at 4:31 PM, Amudhan P <amudhan83 at gmail.com> wrote:

> I mean time takes for listing folders and files? because of "rebalance fix
> layout" was not done.
>
>
> On Tue, Apr 4, 2017 at 1:51 PM, Amudhan P <amudhan83 at gmail.com> wrote:
>
>> Ok, good to hear.
>>
>> will there be any impact in listing folder and files?.
>>
>>
>> On Tue, Apr 4, 2017 at 1:43 PM, Nithya Balachandran <nbalacha at redhat.com>
>> wrote:
>>
>>>
>>>
>>> On 4 April 2017 at 12:33, Amudhan P <amudhan83 at gmail.com> wrote:
>>>
>>>> Hi,
>>>>
>>>> I have a query on rebalancing.
>>>>
>>>> let's consider following is my folder hierarchy.
>>>>
>>>> parent1-fol (parent folder)
>>>>               |_
>>>>                  class-fol-1 ( 1 st level subfolder)
>>>>                                |_
>>>>                                   A ( 2 nd level subfolder)
>>>>                                    |_
>>>>                                       childfol-1 (child folder created
>>>> every time before writing files)
>>>>
>>>>
>>>> Now, I have a running cluster with 3.10.1 with disperse volume and I am
>>>> planning to expand cluster by adding bricks.
>>>>
>>>> will there be a problem using newly added bricks without doing a
>>>> "rebalance fix layout" other than existing files cannot be rebalanced to
>>>> new brick and files created under existing folder will not go to new brick?.
>>>>
>>>> I tested above case in my test setup and observed files created under
>>>> new folder goes to new brick. and I don't see any issue on listing files
>>>> and folder.
>>>>
>>>> so, My case is we create child folder every time before creating files.
>>>>
>>>> The reason to avoid rebalance is I have more than 10000 folders across
>>>> 1080 bricks. so triggering rebalance will take a long time and in my
>>>> previous expansion in 3.7 was not able to access some folders randomly
>>>> until fix layout completes.
>>>>
>>>>
>>> It sounds like you will not need to run a rebalance or fix-layout for
>>> this. It should work fine.
>>>
>>> Regards,
>>> Nithya
>>>
>>>>
>>>> regards
>>>> Amudhan
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Gluster-users mailing list
>>>> Gluster-users at gluster.org
>>>> http://lists.gluster.org/mailman/listinfo/gluster-users
>>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170406/d8a23ea3/attachment.html>