[Gluster-users] Upgrading ancient gluster version

[Dietmar Putz]

just checked my own mail archive...
i don't know if it is supported but a played around with downgrade and i 
was able to downgrade from 3.6.7 to 3.5.6 ...on VM's.
in any case (upgrade or downgrade) you should be aware of the operating 
version...as far as i can remember i had always to set it manually.
this was not done automatically by upgrade or downgrade

http://www.gluster.org/community/documentation/index.php/OperatingVersions


downgrade gluster von 3.6.7 auf 3.5.6 :
situation ... 4 nodes in dist-repl. gfs 3.6.7 + same as geo-rep slave

add-apt-repository ppa:gluster/glusterfs-3.5

gluster volume geo-replication stop <master> <slave>
gluster volume stop <volume>
gluster volume reset <volname> force      (save all settings before...:  
gluster volume info)
apt-get update
apt-get install glusterfs-server=3.5.6-ubuntu1~trusty1 
glusterfs-common=3.5.6-ubuntu1~trusty1 
glusterfs-client=3.5.6-ubuntu1~trusty1
gluster volume set all op-version 30500
gluster volume set <volname> <key> <value> (for all settings)
gluster volume start <volname>

afterwards i have been faced with two errors :

a log file in /var/log/glustserfs/... (forgotten which one...) showed a 
wrong op-version...a 30600 was still in use while 30500 was expected. 
possibly this could be prevented by setting the op-version before 
stopping the volume. i'm not sure. i adapted the 
/var/lib/glusterd/glusterd.info manually...and as far as i can remember 
i was able to start the volume again.


then :
/var/log/glusterfs/bricks/...
2015-12-07 17:59:17.020988] E [graph.y:212:volume_type] 0-parser: Volume 
'myvol-barrier', line 31: type 'features/barrier' is not valid or not 
found on this machine

another :
gluster volume reset <volume> force

sovled the situation.
at least now i was able to start and mount the volume.

however, when you decide to upgrade to 3.6 it's pretty sure that you 
will be faced to the problem that clients cannot mount the volume.
( solved by glusterd --xlator-option *.upgrade=on -N )
https://bugzilla.redhat.com/show_bug.cgi?id=1191176

br
dietmar


Am 12.02.2016 um 14:26 schrieb Dietmar Putz:
> Hi Dave,
>
> based on my experience i would expect problems when old clients trying 
> to access a volume with an upgraded gfs.
> as mentioned recently i upgraded step by step from 3.4.7 to 3.7.6 and 
> as far as i can remember i had problems with a 3.4.7 client when 
> accessing a 3.5 volume. because of the geo-replication problem i was 
> under pressure and have not been focused no such problems...afterwards 
> i upgraded all the clients as recommended before they accessed the 
> volume and i never had problems again.
>
> ...except this little example :
> gluster volume is a 6 node dist.repl. running 14.04 LTS with gfs 
> 3.7.6...the volume is accessed by some 3.7.6 clients and there is 
> still one client with 12.04 LTS and gfs 3.6.7 (3.7 is not available on 
> 14.04 :( so i have to update the entire node...).
> two days ago i tried to set the geo-replication.indexing to off.
>
>
> [ 21:01:49 ] - root at gluster-ger-ber-08 
> /var/lib/misc/glusterfsd/ger-ber-01 $gluster volume set ger-ber-01 
> geo-replication.indexing off
> volume set: failed: Staging failed on gluster-ger-ber-12-int. Error: 
> One or more connected clients cannot support the feature being set. 
> These clients need to be upgraded or disconnected before running this 
> command again
>
> after disconnecting the client with gfs 3.6.7 (the 3.7.6 clients are 
> still connected)  :
>
> [ 21:05:03 ] - root at gluster-ger-ber-08 
> /var/lib/misc/glusterfsd/ger-ber-01 $gluster volume set ger-ber-01 
> geo-replication.indexing off
> volume set: success
> [ 21:12:56 ] - root at gluster-ger-ber-08 
> /var/lib/misc/glusterfsd/ger-ber-01 $
>
> afterwards i was able to connect the 3.6.7 client again.
> i don't know all the features being supported by the different gfs 
> versions but this example should show that they exist.
> a developer could tell you much more about this...
>
> i have to mention that i always followed the 'scheduling a downtime' 
> part like in :
> http://www.gluster.org/community/documentation/index.php/Upgrade_to_3.5
>
> hope that helps...
> best regards
> dietmar
>
>
> On 11.02.2016 23:05, Dave Warren wrote:
>> On 2016-02-11 04:27, Dietmar Putz wrote:
>>> and i strongly believe you have to update all your clients too.
>>> maybe a developer can give you more background information about the 
>>> need to do that...
>>
>> I intend to update all of my clients as soon as possible -- I was 
>> more concerned about whether the old clients will cause any problems 
>> (to the servers, data integrity, etc) until they're upgraded?
>>
>> I'd be perfectly willing to block old clients from connecting (and am 
>> willing to do so from the firewall if needed)
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160212/2d4de830/attachment.html>