[SOLVED] error fetching datastores - 500 after upgrade to 2.2

[Lephisto]

Using at least a Special Device for Metadata solves this. Official recommendation for PBS is still use SSDs. This has not changed. So this is more a less a self made problem for users using rotating disks. The design in PBS and its deduplication needs millions of files and short access times to them.

You know that this is not the reality. No one uses flash as the only Backup Medium, just because PBS suffers from bad technical design decisions. Even today it's just too expensive.

 

[itNGO]

You know that this is not the reality. No one uses flash as the only Backup Medium, just because PBS suffers from bad technical design decisions. Even today it's just too expensive.

Yes I know, but that changes nothing about this is clearly documented as "recommendation". Being below the recommendations and crieing that is slow does help noone. And in Enterprise Datacenter all-flash for backup is nothing special anymore... not at our DCs and not in many others I know.... If you need a low time to recover, there is no other option than having flash.... I talk about getting a 100GB Database Server restored in below 5 minutes cause every minute costs 10k$ when the system is not available.... That PBS design is not considering to be used on rotating disk was known from the very beginning. This does not mean for small deployments it does not work, but don't say the docs haven't warned you....

 

[Sprinterfreak]

Yes I know, but that changes nothing about this is clearly documented as "recommendation". Being below the recommendations and crieing that is slow does help noone. And in Enterprise Datacenter all-flash for backup is nothing special anymore... not at our DCs and not in many others I know.... If you need a low time to recover, there is no other option than having flash.... I talk about getting a 100GB Database Server restored in below 5 minutes cause every minute costs 10k$ when the system is not available.... That PBS design is not considering to be used on rotating disk was known from the very beginning. This does not mean for small deployments it does not work, but don't say the docs haven't warned you....

This issue is all about PBS failing on high load or small deployments.
This issue is also about failing silently for I guess a lot of people who don't have turned on extensive reporting.
Bottom line.

Fine for You @itNGO when You can afford and/or require fast restore times. But this here also has the potential of all Your data falling appart when You suffer from a larger scale recovery, when simply a slight lag occurrs and a useless unhandled timeout, for which we know exists, tips over Your entire backup infrastructure. And yes, this can also hit You on all flash. Just with a slightly less probability.

I don't care about recovery times despite of would be nice to have when money could by. (Which is not the case atm.)
I need a backup system which has all the time of the day to backup approx. 10 VM's total, which practically has a daily runtime of 4 minutes. All fine. Sadly a third of backup tasks per job just don't start because they run into a fairly short unhandled timeout after a few seconds. I would prefer if the whole backup just took a minute longer instead of dropping a random third of all data for no good reason.