[INSTRM-389] Add cache='none' to virt xml Created: 21/Jun/18  Updated: 14/Mar/23  Resolved: 14/Mar/23

Status: Won't Fix
Project: Instrument control development
Component/s: ics_production
Affects Version/s: None
Fix Version/s: None

Type: Task Priority: Normal
Reporter: shimono Assignee: Yoshida, Hiroshige
Resolution: Won't Fix Votes: 0
Labels: subaru-personnel
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified


 Description   

To be capable with hot migration at the production, add cache='none' to locally attached storage.

also refer subaruif #430 at https://pfs.ipmu.jp/pipermail/subaruif-software/2018q2/000429.html



 Comments   
Comment by kyono [ 23/Jun/18 ]

Thank you for your information. We will test this on Summit system with one of lf vms.
Once we finish it, will update status here.

 

Comment by kyono [ 27/Jun/18 ]

shimono san

We have tested hot migration, but it was not successful...

[Status]

  • From vm host perspective, it seemed like a vm guest was moved from one host to another.
      (i.e. with virsh list --all)
  • After hot migration, there was no network connectivity to the vm guest.
      (I assume this is because of learned vm guest's MAC address stayed on the same port on pfs core and switch not being able to handle L2 communication...)

>- service connectability during live migration has not tested
>
Does this one mean what I see now?

I will do a little bit more research, but could you let us know the hardware connection among IPMU servers and network devices you tested hot migration?

P.S
This is just comment for [PFS-Soft-Subaru:423] . It might be obvious, but we need to redefine VM guest as we modify .xml file.

Again thank you for sharing this info, shimono-san.

 

Comment by kyono [ 27/Jun/18 ]

oh,,,
never mind regarding MAC address...

Comment by shimono [ 27/Jun/18 ]

`service connectability during live migration` is on a VM guest moving from one to another, not before migration started nor after migration finished.

target VM guests has two configurations, one is only with one network (PFS), another is with two networks (PFS, IPMU). PFS has two switches, cisco 2970 and fujitsu SR-S324, IPMU is by DELL 5426.
VM hosts on PFS network was connected through two switches, like hostA - 2970 (lacp) SR-S324 - hostB.

if you could not solve issue, please print `virsh dumpxml` and `brctl show` (here or subaruif ml for closed).

Comment by yuki.moritani [ 14/Mar/23 ]

No demand now. Note that Yoshida-san has been hot-migration of a few servers at the summit.

Generated at Sat Feb 10 16:24:30 JST 2024 using Jira 8.3.4#803005-sha1:1f96e09b3c60279a408a2ae47be3c745f571388b.