LAGG/LACP Slow timeout
-
Hello All,
I am currently in the middle of testing a psfense cluster for possible use in my group. I my testing I have come across and issue with the pfsense/bsd implementation of LACP. If a link in the LAGG on the pfsense goes down for what ever reason I lose connectivity for about 5-6 secs. My traffic cannot have this long of a drop without serious issue. After investigating, it appears that pfsense/bsd defaults to a setting of slow timeout which can take up top 30 secs for LACP to converge. This is too long but I cannot find any documentation how to change this setting. The lacp standard allows for two types of timeout fast and slow. I would like to set this to fast which is a 1 sec timeout. Does anyone have any idea of how to set this? I found this link https://www.freebsd.org/cgi/man.cgi?query=lagg(4) and see that at the end there are no way to se admin options but though I would see if anyone has seen this and found a work around or an older setting.
Thanks for any input,
jpagan
CCNP, JNCIP -
I my testing I have come across and issue with the pfsense/bsd implementation of LACP.
In front of the cluster or behind it, to the LAN or WAN side I mean?
Do we talking about dynamic LAG over LACP and automatic set up or do we talking about static LAG manual set up?
active/passive or active/activeIf a link in the LAGG on the pfsense goes down for what ever reason I lose connectivity for about 5-6 secs.
To what the LAG is connected? Is it a Switch and if so what kind of switch? (vendor, model, number)
LAGs are from 2 till 8 links, how many do you use?My traffic cannot have this long of a drop without serious issue.
- Is there an fail over configuration or a load balancing set up?
- If so, why do you not using then 10 GBit/s links?
After investigating, it appears that pfsense/bsd defaults to a setting of slow timeout which can take up top 30 secs for LACP to converge.
I haven´t until now seen or heard about 30 seconds to take over.
In normal if this LAG is a dynamic one and uses the LACP, and you set up load balancing, that
means all 2 till 8 links are up and working together likes one big fat pipe and then one link is
failing you will have then 10/100/1000 or whatever the link speed is, less, but you will not really
know or see it!!! Let us imagine you are using 8 GB LAN Ports together, to aggregate one fat pipe
acting as one 8 Gbit/s link and then one link is failing and you only owns a LAG with 7 GBit/s that
is active running, you really don´t will see any bottleneck in your network. -
@BlueKobold:
I my testing I have come across and issue with the pfsense/bsd implementation of LACP.
In front of the cluster or behind it, to the LAN or WAN side I mean?
Do we talking about dynamic LAG over LACP and automatic set up or do we talking about static LAG manual set up?
active/passive or active/activeonly dynamic lacp sends the fast/slow pdus, so must be dynamic.