PBS Sync bandwidth very slow over Wireguard

[garnoux]

Hi,
i am currently experiencing a problem with a PBS environment located on two remote sites.
Both sites are linked via vpn tunnel because there is no fixed ip address on the second, and therefore it is impossible to nat port 8007 limited to a source address
so i have a wireguard vpn tunnel set up between two sites via pfsense (physical appliances)

but after launching a sync job, i notice that the speed is around 100kbps. impossible to understand why

here are the different tests carried out

speedtest site 1

speedtest site 2

iperf site2 to site1 via wan

iperf site2 to site1 via wireguard

(slow but okey)
proxmox benchmark via wan

proxmox benchmark via wireguard

here is the actual job logs

2024-08-13T17:21:26+02:00: Starting datastore sync job 'pbs-remote:backup:DATA::s-acf37434-96dc'
2024-08-13T17:21:26+02:00: sync datastore 'DATA' from 'pbs-remote/backup'
2024-08-13T17:21:26+02:00: ----
2024-08-13T17:21:26+02:00: Syncing datastore 'backup', namespace 'millery' into datastore 'DATA', root namespace
2024-08-13T17:21:27+02:00: found 13 groups to sync (out of 13 total)
2024-08-13T17:21:27+02:00: skipped: 1 snapshot(s) (2024-07-27T23:12:48Z) - older than the newest local snapshot
2024-08-13T17:21:27+02:00: re-sync snapshot vm/30001/2024-07-31T22:00:05Z
2024-08-13T17:21:27+02:00: no data changes
2024-08-13T17:21:27+02:00: percentage done: 7.69% (1/13 groups)
2024-08-13T17:21:27+02:00: re-sync snapshot vm/6155/2024-07-31T22:00:05Z
2024-08-13T17:21:27+02:00: no data changes
2024-08-13T17:21:27+02:00: percentage done: 15.38% (2/13 groups)
2024-08-13T17:21:34+02:00: skipped: 38 snapshot(s) (2024-02-29T21:00:02Z .. 2024-08-02T23:00:03Z) - older than the newest local snapshot
2024-08-13T17:21:34+02:00: skipped: 2 snapshot(s) (2024-08-04T23:00:03Z .. 2024-08-05T23:00:00Z) - due to transfer-last
2024-08-13T17:21:34+02:00: re-sync snapshot vm/8009/2024-08-03T23:00:01Z
2024-08-13T17:21:34+02:00: no data changes
2024-08-13T17:21:34+02:00: percentage done: 16.35% (2/13 groups, 1/8 snapshots in group #3)
2024-08-13T17:21:36+02:00: re-sync snapshot vm/8009/2024-08-06T23:00:02Z
2024-08-13T17:21:36+02:00: sync archive drive-virtio0.img.fidx

and the network graph



i tried to lower the mtu a little but that doesn't change anything
could you clarify this for me?
thanks

 

[patrick.pesegodinski]

What is the latency between the two sites?

 

[garnoux]

Sites are 10km-distant.

 

[garnoux]

Hello, any up on this subject ?

 

[_gabriel]

are the two sites has no fixed ip ?

 

[garnoux]

Site 1 has fixed, site 2 has Dynamic. But changes are very rare.

 

[_gabriel]

so site1 pull data from site2 ?
= site2 is the remote peer in site1 PBS, right ?

 

[garnoux]

No, site 2 pull from site one
Using wireguard because i cant open 8007 port restricted by the site 2 IP (Dynamic)

 

[_gabriel]

here I set a dyndns to site2, then its ip in an ipset of site1 pve firewall.
+ crontab to update the ipset ip.

 

[garnoux]

but why i have this bandwidth on wireguard ?

 

[Darkk]

Might be something to do with the MTU on Wireguard. Wireguard generally don't have any restrictions on bandwidth. Wireguard will use much bandwidth as you give it. Sometimes ISPs mangle the MTUs on their network which result in poor performance. Some protocols are sensitive to that.