R81 VSX

[genisis__]

I was wondering if anyone has actually deployed R81 in a VSX setup yet and if this has any  reported issues? 

I'm looking to upgrade my R80.20 VSX setup to R80.40 or R81, would like to move to R81 but I think it may be a little too early for this.

Also I know the recommendation is to rebuild, but in the current climate remote upgrade is preferred method.  So I will likely do an inline upgrade; from what I can tell kernel version would get upgraded, multi-queue turned on and other parameters turned on by default such are CORE load balancing parameter (SK168513).

Clearly new filesystem would not get used, but I don't see this being hugely important at the gateway side (happy to be educated on this if I'm wrong).

[genisis__]

If we have dynamic balancing enabled, how does this affect monitoring fwk process load on a per VS basis?  At the moment we use SNMP to determine the number of fwk processes assigned to a VS; with dynamic balancing turned on I can this this ability to historically track utilisation at a VS level will break.

[AmitShmuel]

I'm not sure if I understood correctly, but Dynamic Balancing doesn't change the number of fwk processes/threads, it simply changes the set of cores they are affined to.

[genisis__]

ok - may sounds silly, but it would be good to see this being monitored then load add and watch the process.

So what your saying is potentially I could have a single core assigned to a VS (example), and if the load increases dynamic balancing will may add additional cores to that fwk process?

Also with TP blades such as IPS, and the bypass threshold being 70% (default) I assume the decision to add/remove cores is also aligned i.e. the CPU utilisations of a a VS /FWK process should not go over 70% in order to ensure blades are not bypassed due to load.

[AmitShmuel]

let's take an 8 cores (no HTT) machine example:

cores 0-1 for SNDS
cores 2-7 for FWKs (of all VSs)

If SNDs are working harder, the split will change to:

cores 0-2 for SNDS
cores 3-7 for FWKs (of all VSs)

That of course, will balance the load, reducing SNDs', and increasing FWKs'.

Dynamic Balancing ensures FWKs load will not pass the 50% threshold, given the SNDs are not twice more loaded, that is of course configurable.

[Massimo_Manzato]

Many thanks

[Alex-]

Are there more R81 VSX deployments out there? I've been upgrading a cluster of 26000T who ran the out-of-the-box R80.30 to R81 Take 44.

The first machine ran the in-place CPUSE upgrade flawlessly, but after a few hours of production time, the root partition filled up to 100% with plenty of directories and files appearing all over the place in the CTX directories I don't see on other systems still running R80.40. SR open. In all fairness, this issue began to appear on r80.30 and I hoped the upgrade would solve it.

The second unit didn't succeed with the in-place upgrade, with messages the configuration can't be imported. I did a fresh install from USB, vsx_util reconfigure, but I ran into sk105441. The SK indicates the issue is fixed in R81, so I left feedback in the SK as it's apparently not the case. Once this was addressed, the very final step of vsx_util reconfigure fails because of some communication issue. I had a productive session with TAC to get more logs for analysis. 

I'm not sure now if I should have gone for R80.40 with the VSX. I've upgraded appliances running classical FW to R81 T44 and no issues there.

[genisis__]

I did a R80.30 to R80.40 'clean' build with the latest Jumbo at the time and this went fine.  There was an issue with in-place upgrade which was a nightmare and had to do a full rollback.

I've not gone to R81.x yet, but I would certainly feel like a clean install is the way to go.  Personally I would probably go straight to R81.10 with JHFA9 or greater.

And most importantly ensure you have a pro-active case raised and get TAC on a zoom session with you (not some 1st liner but someone experienced with VSX)