Backups stop working.

[iccbroadcast]

Hi!

Our Backups have stopped working.
We have a 5 node cluster running, v 6.2 , going to its third year now (so far, so good...its reaching its planned hardware EOL, so we will replace by new one soon), but recently, it has been restarted.

After coming up, some back-ups does not work any more... I think all of the ones using NFS storage.

Error message on Backups says: cannot have 'maxfiles' and 'prune-backups' configured at the same time

Maxfiles was initially not set up at GUI Datacenter/storage entries, but it was marked as a required field.
Initially, backups were not done, because the backup system complained that there too mach backups in place, and asked to delete some to make room.
Now but, after 'normalizing' the GUI situation (getting rid of the red form field, since the field seems to be mandatory), backups do not work anymore.

I have tried removing and re-creating the backup task in GUI, but that didn't help.
I'm thinking on removing and re-declaring the NFS storages, but, considering this has been working for years, maybe it is better to assess the situation with the help of some expert.

Cheers!

 

[iccbroadcast]

Hi... It seems that, yet another time, version mismatch among cluster members may be the culprit.

Still not sure, but I guess it is, since other cluster members have successfully done backups.

The problem is that, updating the oldest node will put it 'ahead' of the rest... and who knows what surprises my arise.

It would be nice if it could be possible, for a new member joining a cluster, to match the rest of the cluster version/packages, even if that meant downgrading to those packages.

 

[Neobin]

We have a 5 node cluster running, v 6.2 , going to its third year now (so far, so good...its reaching its planned hardware EOL, so we will replace by new one soon)

Hi... It seems that, yet another time, version mismatch among cluster members may be the culprit.

Still not sure, but I guess it is, since other cluster members have successfully done backups.

The problem is that, updating the oldest node will put it 'ahead' of the rest... and who knows what surprises my arise.

It would be nice if it could be possible, for a new member joining a cluster, to match the rest of the cluster version/packages, even if that meant downgrading to those packages.

Setting up a cluster and then never updating it for its whole lifetime is a really bad practice, in my opinion.

One of the pros of a (well dimensioned) cluster in combination with VMs, is the ability to evacuate a node, update and reboot it, without any interruption of the guests / the services inside the guests. (Special cases with host-bound guests, e.g.: PCIe-passthrough and alike, aside.)

If you keep the whole cluster up-to-date on a regular basis, there is no need to downgrade anything.

The goal should ever be, that all nodes in a cluster are on the same recent version.

Just my 2 cents...