Sync causing problem in our wan

[Elleni]

Hi all - first of all I want to express a big thank you for your great products pve and pbs which serve us very well. Keep up the impressive work.

While testing a multiple pbs setup for our company we discovered a severe problem with one of our sites, which presented itself while syncing datastores from one pbs to another. This happened despite me having calculated and limited the sync job to approx. 50% of the available bandwith.

What happend is that on the remote site where we pushed a sync to, the dhcp server (windows server) started to have problems. We have multiple subnets on the remote site, and while the subnet on which the dhcp server is itself member of still is operational - meaning it offered ips to dhcp requests, on the other subnets where the dhcp server offers dhcp service via configured helpers all of a sudden it stopped offering answers thus dhcp clients did not get ip's anymore until I stopped the sync job. The pbs server lives in one of those subnets - served dhcp services with helpers.

The only thing I changed on the pbs setup - and I dont know if this could have an impact - is the blocksize which I set to 4MB, which in the meantime I doubt myself if it was worth to set anyways, as garbage collect is reporting an average chunk size of 1.672 MiB.

Does anyone have experienced similar problems and has an idea on how to troubleshoot this please? Thanks in advance - please let me know what informations you might need to further analyze this.

 

[Elleni]

No-one an idea, what could cause this, or howto troubleshoot this further?

 

[Johannes S]

How is your network setup? Do you have a dedicated connection for the backup or is all running over the same network link? You can set a rate limit in the configuration of the sync job, did you try that?

 

[Elleni]

Hello Johannes,

our company has different sites - connected by a dedicated 500 Mbit Wan provided by our internetprovider. No there is no dedicated connection for backup, so its running on the same network link. As written in the opening post I already had set a ratelimit of about half the bandwith to not interfere but the problem arise anyway. Thats why I thought to ask here as maybe someone else already has seen that symptom and could know where it comes from and how it can be fixed.

 

[Johannes S]

Sorry, I missed that line about the set limitation. I have honestly no idea how to proceed, but I would try to limit the bandwith for the sync job even more to an absurd low amount just to test whether this change anything. I would expect that with a rather small amount (like 10%, 5% or even 1% of the bandwith) at some point the DHCP issues won't arise anymore. If they still arise the question would be whether the problem is actually on PBS or whether you could trigger the same problem by setting up a huge data transfer via http/2 (which is used by PBS afik) or http/https with a similiar amount of data like your backup jobs.
Another shot: Do you happen to have any VPN envolved in the setup? I'm wondering whether that might be a bottle neck (although this is more guessing than anything else so take this with a grain of salt).