CDP Optimization

[halllowien]

Hello,

we have 350 Machines covered by CDP.

As we can see the Most IOPS \ MB/s occur on the moment the Short-time Retention delete.

Is it Possible to Change the Start of this for different CDP-Jobs?
We have 15 Jobs with a Short-time Retention of 5 Minutes.

Greets
David

[HannesK]

Hello,
the goal of the policy is to satisfy whatever RPO is configured. So there is no "start time". As you have only 5min retention, I guess that you only have a few seconds RPO time. So I'm not sure, how different timings by a few seconds could help.

Overall I guess, that the question is related to performance issues. You could open a Veeam support case and check for performance issues (prio 4 case). To give support a chance to say something, they also need details on the hardware. In best case, an infrastructure diagram / picture with all the components involved. Please post the case number for reference, then I will also have a look at it.

Best regards,
Hannes

[halllowien]

Hello,

in veeam one i can see that there is high io peeks on the same time for all Jobs.
for me it looks like there is a scheduler for the retention, because i have more than this 5 minutes of short-time retention.

Greets
David

[HannesK]

Hello,
yes, everything that has the same retention will be executed at the same time. So the behavior sounds expected. That peak can be avoided by setting different retentions (e.g. 5min, 6min, 7min retention on different jobs).

We have 15 Jobs with a Short-time Retention of 5 Minutes.

because i have more than this 5 minutes of short-time retention.

now I'm confused Is it maybe that you have 5min RPO time and a longer short-term retention? If yes, what is the short-term retention and is it the same on all policies?

Just curious, If you really have 5min short term retention: what is the purpose of keeping it that low?

Thanks,
Hannes

[halllowien]

Hello,

we have not enough space for much more Short Retention.

All Our Jobs have 60 Seconds RPO , 5 Minutes Short time retention, 24 Hours Longterm retention with 2 days.

On failover a VM there are around 10 minutes of Short retention points to Choose.

With this and the veeamone performance logs i think that Veeam CDP don't delete the configured retention immediately, instead it maybe happen with a scheduler.

Greets
David

[HannesK]

Hello,
without going too much into details: developers recommended to increase the short term retention (as alternative to the 5,6,7 etc. I mentioned above). That makes the short deletion process overall more efficient and reduces the spikes (they recommended 1h for example)

Still curious: what is the purpose of keeping short term retention it that low? I mean, if something goes wrong, are you able to react within 5-10min?

Best regards,
Hannes

[halllowien]

Hello,

we use CDP only as a disaster scenario and not for "let us go 30 minutes back in time".

In this scenario the source Datacenter is lost.

We then start DNS + DC + VBK server from Veeam Replica, disable the CDP Jobs and start Machines in disaster.

if CDP starts to work as expected (we are near on that point) we will get a new San or an upgrade for this and then we can set Short-term to a higher value.

Greets
David