[Gluster-users] What I noticed while upgrading 3.7.6 to 3.7.8

[Atin Mukherjee]

On 02/28/2016 11:06 PM, ML mail wrote:
> Thanks for explaining why the ping-timeout option has been disabled. Now in my case of a small replicate volume of 2 physical nodes with one brick per node, does it mean that if one node goes down the FUSE client mount will not respond at all? In the past it would simply block for 42 seconds and after this timeout respond again.
No, that still applies. ping timer logic still holds true for connection
between client & bricks and by default it should be 42 seconds. Only its
disabled between GlusterD to GlusterD connection.

> 
> By the way is there a ChangeLog somewhere for 3.7.8?
Please refer [1]

[1]
https://github.com/gluster/glusterfs/blob/release-3.7/doc/release-notes/3.7.8.md
> 
> Regards
> ML
> 
> 
> 
> 
> On Sunday, February 28, 2016 5:50 PM, Atin Mukherjee <amukherj at redhat.com> wrote:
> 
> 
> On 02/28/2016 04:48 PM, ML mail wrote:
>> Hi,
>>
>> I just upgraded from 3.7.6 to 3.7.8 and noticed the following 2 points:
>>
>> 1) As others on the mailing list I am also affected by a massive performance drop on my FUSE mount. I used to have around 10 MB/s transfer, which is already quite slow, but now since the upgrade I am at around 2 MB/s as you can see below:
>>
>>
>> $ sudo dd if=/dev/zero of=bwtest bs=1M count=64
>> 64+0 records in
>> 64+0 records out
>> 67108864 bytes (67 MB) copied, 37.9311 s, 1.8 MB/s
>>
>>
>> I know there is a bug (#1309462) already opened for that but just thought I would mention it too and hope that it will be fixed in 3.7.9.
>>
>>
>> 2) I noticed that the ping-timeout option in /etc/glusterfs/glusterd.vol changed from 30 seconds to 0 seconds. Can someone comment on that change? and does this mean that with 0 seconds it never times out?
> This indicates that there is no ping mechanism between GlusterDs now. We
> have had some issues from snapshot use cases when we tried to enable
> ping timer in glusterd, the solution to it was multi threaded epoll and
> when we enabled it then we got to see some corruption and crashes which
> are yet to be RCAed. So we decided to turn both these features off till
> we get to the root cause.
> 
> HTH,
> Atin
> 
>>
>> Regards
>> ML
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://www.gluster.org/mailman/listinfo/gluster-users
> 
>>
>