[om-infra] [om-workshop] 502 Bad Gateway

Jean-Claude Vanier jclvanier at gmail.com
Fri Dec 16 07:24:52 EST 2016


Finally, after having read tons of docs, I got this conclusion :
-- the first type of VM we used is vmdk (in simple words, this isn't
an image but a chrooted system). Jasper is in this case.
-- proxmox cannot do shnapshot with this type of VM
-- proxmox can do snapshots with appropriate types like qcow2
-- we have only one VM of this type: quartz (discourse)
==> snapshots can be done, and actually are already done, for quartz
but not for the other virtual hosts

What to do ?
-- we could continue like this. I changed the the type of compression
for the backups: LZO which is said faster than gzip (which is better).
I'm afraid that will not change the situation fundamentally.
-- we could convert vmdk to qcow2 machines. In such case, some tests
must be done before the real migration.




2016-12-16 1:03 GMT+01:00 Jean-Claude Vanier <jclvanier at gmail.com>:
> Yes, it's almost fixed. I still have to change the behaviour of the
> backup. Instead of suspending the VM, before copying the files, it
> might be more efficient to create a snapshot. For that, it is
> necessary to tweak the config files of proxmox, our VM manager.
> Raphaël, thought about another solution, totally different: for each
> CMS we use, we could rent a very low cost shared VM at online or
> scaleway for example. If we do the things carefully, the total cost
> might be lower than the current one.
> The advantage is that each machine has its own DB and its own IP and
> therefore its own bandwith. Furthermore, the maintenance of the web
> server is no longer our problem. In addition, it could be possible to
> have two low charge CMS in the same host. This kind of architecture
> has to be well prepared, as you can imagine.
>
>
> 2016-12-15 18:52 GMT+01:00 Ben Bullard <benbullard79 at cox.net>:
>> Then it's most likely fixed. I'll let you know. Seems it likes to go down
>> when it's early evening here (Louisiana, US).
>>
>> Ben Bullard
>> ben79
>> --------------------
>> OpenMandriva-QA Team
>>
>> On 12/14/2016 12:28 PM, Jean-Claude Vanier wrote:
>>
>> I think I have found what happened. A cron script was launched
>> regularly, as it must be, but never ended. The system ran out of
>> memory in both ram and disk.
>> I have done some cleaning. I hope all will be right, now.
>>
>>
>> 2016-12-14 8:38 GMT+01:00 Jean-Claude Vanier <jclvanier at gmail.com>:
>>
>> Hi Ben,
>> the server was down for 7 hours this night. We are working on it to
>> find out the reason. It seems that we have a problem with the cache.
>> More infos later.
>>
>>
>> 2016-12-14 1:54 GMT+01:00 Ben Bullard via OM-Workshop
>> <om-workshop at ml.openmandriva.org>:
>>
>> I'm needing to access Release Notes but it's 502 Bad Gateway.
>>
>> https://wiki.openmandriva.org/en/2014.2/Release_Notes
>>
>> https://wiki.openmandriva.org/en/3.0/Release_Notes
>>
>>
>> --
>>
>> Ben Bullard
>> ben79
>> --------------------
>> OpenMandriva-QA Team
>>
>>
>> _______________________________________________
>> OM-Workshop mailing list
>> OM-Workshop at ml.openmandriva.org
>> http://ml.openmandriva.org/mailman/listinfo/om-workshop_ml.openmandriva.org
>>
>>



More information about the OM-Infra mailing list