Gluster users - Apr 2017 - Gluster 3.8.10 rebalance VMs corruption

2017-04-27 13:31 GMT+02:00 Pranith Kumar Karampuri <pkarampu at
redhat.com>:
> But even after that fix, it is still leading to pause. And these are the
two
> updates on what the developers are doing as per my understanding. So that
> workflow is not stable yet IMO.
So, even after that fix, two more critical bug leading to
dataloss/corruption where found ?

I'm sorry but this is pure madness, plese put a "beta" label on
sharding or gluster itself. 4 bugs in a row making loose of data in a
software enginereed to keep data safe are unacceptable.
Yes, all software has bug, but not with this frequency, for the same
reason, with the same results: data loss.

I'm really, really, really warried to put my company valuable data in
a gluster storage. If I loose hundreds of VMs at once, i'm really
fucked.
Yes, backup exists, but try to say to your customers that you have to
restore 20TB from backups (it takes days....) and that they lost many
ecommerce orders/transactions.

On Thu, Apr 27, 2017 at 5:21 PM, Gandalf Corvotempesta <
gandalf.corvotempesta at gmail.com> wrote:
> 2017-04-27 13:31 GMT+02:00 Pranith Kumar Karampuri <pkarampu at
redhat.com>:
> > But even after that fix, it is still leading to pause. And these are
the
> two
> > updates on what the developers are doing as per my understanding. So
that
> > workflow is not stable yet IMO.
>
> So, even after that fix, two more critical bug leading to
> dataloss/corruption where found ?
>
yes

> I'm sorry but this is pure madness, plese put a "beta" label
on
> sharding or gluster itself. 4 bugs in a row making loose of data in a
> software enginereed to keep data safe are unacceptable.
> Yes, all software has bug, but not with this frequency, for the same
> reason, with the same results: data loss.
>
The bugs are not in sharding. Sharding + VM workload is exposing bugs are
in DHT/rebalance. These bugs existed for years. They are coming to the fore
only now. It proves to be very difficult to recreate these bugs in other
environments. We are very transparent about this fact. Until these issues
are fixed please don't do rebalance in shard+VM environments.

>
> I'm really, really, really warried to put my company valuable data in
> a gluster storage. If I loose hundreds of VMs at once, i'm really
> fucked.
> Yes, backup exists, but try to say to your customers that you have to
> restore 20TB from backups (it takes days....) and that they lost many
> ecommerce orders/transactions.
>


-- 
Pranith
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20170427/68c159ba/attachment.html>