[Gluster-users] upgrade 2.0.2 to 2.0.9?

[Count Zero]

I certainly hope you get an answer, as this has happened to me as well in the past (although on a smaller scale). I will follow this thread closely in the hope you get an answer with interesting information.


On Apr 18, 2010, at 9:40 AM, Robert Minvielle wrote:

> 
> I have a test cluster running 2.0.9, I am happy with that and it works. 
> The clients all have correct settings... moving the data is not really
> an option. We are talking about 27+TB here. I copied it to the test cluster
> as a backup, that took 4 days. I don't have enough room on each brick to
> do a /data/export.old. 
> 
> Funny thing is, I did this on the test cluster and it worked. I recall
> the gluster team had told me to do something different back then, but
> I don't have that email. Again, each bricks export is over 50%, so I
> can not make a copy and then copy it back. Copying it all to the backup
> cluster and back would take over a week. Even a sync takes three days,
> so I am kindof in a jam to figure this out without moving data. 
> 
> 
> 
> ----- "Count Zero" <countz at gmail.com> wrote:
> 
>> Make sure all your clients know about all your bricks, make sure the
>> hostnames are correct, etc.
>> Make sure there's communications between the nodes, on the glusterfs
>> prort.
>> And at worst, move the /data/export directory to /data/export.old and
>> make a new /data/export
>> Remount on all servers, and run some tests (perhaps with bonnie++).
>> 
>> Once you are happy with the performance and functionality, move the
>> files in /data/export.old/ back to /data/export by copying them over.
>> 
>> 
>> 
>> On Apr 18, 2010, at 9:25 AM, Robert Minvielle wrote:
>> 
>>> Yes, tried that. Same thing. 
>>> 
>>> I thought I recalled that would fix it... but it seems to not
>>> work. 
>>> 
>>> ----- "Count Zero" <countz at gmail.com> wrote:
>>> 
>>>> If I remember correctly, the suggestion was to type: ls -lR
>>>> (You do this from your mounted directory).
>>>> 
>>>> In theory, this forces glusterfs to update/heal/etc.
>>>> 
>>>> Aric
>>>> 
>>>> 
>>>> On Apr 18, 2010, at 9:07 AM, Robert Minvielle wrote:
>>>> 
>>>>> 
>>>>> I have two glusters, a production and test gluster. A while back,
>> I
>>>> upgraded the
>>>>> test gluster from 2.0.2 to 2.0.9 with no problems. Both glusters
>> are
>>>> running gentoo,
>>>>> same version, same kernel. The client saw all of the files,
>>>> everything worked.
>>>>> 
>>>>> Over this weekend I upgraded the main production gluster, as we
>> are
>>>> trying to get some
>>>>> more speed and I was told by the gluster team some time ago to
>> try
>>>> out 2.0.9 to cure
>>>>> a few problems. I did the same exact upgrade that I did on the
>> test
>>>> gluster.
>>>>> Now that I am done, I can mount it fine and I get no errors in
>> the
>>>> logs, however,
>>>>> most (e.g. 85%) of my files are missing, or I get a "cannot
>> access
>>>> /data/blah/blah no
>>>>> such file or directory". The files exist on the server in the
>>>> exported directory,
>>>>> but the client does not see some of them. I recall there was
>>>> something I had to do
>>>>> to fix this perhaps... I can not recall what it is.
>>>>> 
>>>>> Anyone have any ideas? (other than shutting down work tomorrow,
>>>> ugh)
>>>>> 
>>>>> At this time I may be forced to back off to 2.0.0... hope that
>> works
>>>> :O
>>>>> _______________________________________________
>>>>> Gluster-users mailing list
>>>>> Gluster-users at gluster.org
>>>>> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users