https://worldofeuc.com logo
Join Slack
Powered by
This message was deleted.
# citrix-vad
s
This message was deleted.
w
as low as you can afford
👍 2
r
i swore i read somewhere that like 5ms and under was "best practice", but i can't seem to find the reference
l
You're better off looking at this from a hypervisor POV to be honest. Anything above 5ms is going to be impacting though. I'd say you have a good steer.
r
👍
l
I went back on some notes I made. Spikes of 20ms can be seen on mechanical storage and was only (back in the day), 3-4ms spikes with flash. This should not be constant though.
r
Thank you for the followup info
j
thanks Ryan for posting this. Guys I was asking Ryan about this because we run all of our customer file servers on Nutanix with FSLogix and consistently seeing 10-25ms on reads. writes are never much more than 5ms. trying to determine if thats a major issue in our environment
l
@Kees Baggerman you need the man that can. Where are these metrics coming from? The VM? With Nutanix I understand the replication piece and how its displaying files around but I cannot comment on what is good or bad from a HCI standpoint.
r
Good point. I bet Kees can chime in here
l
Absolutely. He's always really helpful, a pretty nice guy too ;-)
r
he's probably lurking in #CLMBQP3PA
or, he's at the pub. probably need to wait until tomorrow for him to chime in
l
Ha ha, lurking... made me laugh.
r
or @Jarian Gibson also...
l
Aka the thread wipper lol , as he always tells me off for not using threads ;-)
k
Sorry guys, I was actually doing work related stuff 🙂
Where is the latency measured?
j
using LogicMonitor so hypervisor metrics
k
Ok.. But is it the read access of the FileServer VM (FSVM) to the VG or is it the read access of the User VM (UVM) to the FSVM?
i.e. if you go to prism and you look at the UVM, do you see read latency there or do you see the latency on the FSVMs?
r
work related stuff? isn't it like 9pm where you are?
j
in prism, I can see the high read latency. like currently I can see a point in time metric of 69.22ms on this particular virtual disk where the VHDs live
k
9:25PM, running CDF traces to show Citrix they’re doing different power management calls to AHV if you compare StoreFront to Studio
j
I think thats where LogicMonitor pulls its data, so obviously this is a point in time metric, but the LogicMonitor average is around 15-20ms
^^whoa. good to know! but scary
k
Ah that’s read latency from the FSVM, I don’t think this is worrying. It’s a VG construct so slightly different from the typical VM vDisk.
j
ah ok...it does seem that especially on Monday mornings for logins, it feels that we have some tiering issues, where the VMs, even in guest, are slow to respond. Once we get some good activity on that machine, it starts running properly. I dont have any definitive data to say thats whats happening, but we did have that issue in the past. had Nutanix support adjust some gflags a few times, but I think most of those have been reverted to default with some of the later updates
k
Are you doing a-sync replication at any point? Cerebro might be causing spikes during diff scanning?
https://portal.nutanix.com/#/page/kbs/details?targetId=kA00e000000CrUKCA0 that’s another potential cause and is inline with your comments on monday
j
ok, looks like that link is no good for me, but Ill send it to a coworker that may have more access in the nutanix portal. Also, talking with him, it sounds like Logicmonitor polls in-guest numbers. looking at Prism historical, the numbers are definitely higher in prism, but the LogicMonitor numbers are still in the 15-20+ms range
no a-sync replications going on
k
Have a look at the KB, if it doesn’t apply and you’re still not comfortable call in support. Apparently they do an amazing job ;-)
j
haha. sounds good. really appreciate your time!
l
You rock @Kees Baggerman
k
We all play our little part @Leee Jeffries 👍🏻
l
Aye we do. Someone at your level in an organisation taking time to chip in on something like this is rare.
w
It is normal for @Kees Baggerman and @Jarian Gibson
l
I think it's very cool
j
Just catching up on this. Looks like Kees has it handled.
j
@Kees Baggerman sorry, but the link isnt working. could you send me the KB article number? many thanks!
j
8157
k
I just put away my laptop, I can get you... or Jarian can get it ;-)
j
yeah no worries. thanks @Jarian Gibson. Looks like our org doesnt have access to that one for some reason. if you can send me the content in a DM or email?
j
Ah it’s marked internal. I just checked it on my phone. Will when I get back from picking kids up from school.
j
awesome. thanks again!
hey @Jarian Gibson, would you mind sharing that KB when you get some time? Thanks!
j
Yes DM coming
👍 1
Open in Slack
PreviousNext
Powered by