Re: Reliable WAL file shipping over unreliable network

On 01/03/18 18:18, Rui DeSousa wrote:

> I can tell you this; it’s not ZFS or it’s quotas.  I know that Postgres has no problems with ZFS and quotas; I have
seenPostgres development environments run out of quota space on ZFS and Postgres never got corrupted — it would just
crashand end up in a crash/restore loop.  Just increase the quota and Postgres recovers without issue exiting the
crash/restoreloop, cleanup the rouge space usage, and reduce the quota back down.  It makes recovering from out of
spaceissues very quick.
 
>
>

So enabling quotas on the same ext4 filesystem I reliably getting 'quota 
exceeded' and 11 return code from rsync. This is on Ubuntu 17.10 (not 
sure what you are on)...but to me it points the finger at a ZFS bug.

Also as has just been pointed out - typically people enable monitoring 
and provision so as to specifically avoid the weird edge cases for out 
of disk/quota etc.

regards
Mark