vm migration failing


Userlevel 1
Badge +10
Perhaps this is tied to my disk replacement adventures, but the host with the failed disk, after being put back into service, doesn't seem to allow host migration. I can migrate just fine between two of my three hosts, but to the third one, I get:

Operation failed: Host *ID* is not schedulable

This topic has been closed for comments

9 replies

Badge +10
I can not migrate a vm from one node to another. It fails with

"Error while migrating vm d0968db8-f45c-4e22-916b-7bf70f0d8a9d: internal error: received hangup / error event on socket".

Using acli shows this:

vm.migrate vm1
vm1: kMigrationFail: Error while migrating vm d0968db8-f45c-4e22-916b-7bf70f0d8a9d: internal error: rece[...]
----- vm1 -----
kMigrationFail: Error while migrating vm d0968db8-f45c-4e22-916b-7bf70f0d8a9d: internal error: received hangup / error event on socket


There is following warning everywhere in logs/acropolis.log:

2015-08-18 09:45:08 WARNING arithmos_mixin.py:231 Could not read dir /sys/class/net on d8185a67-e3b6-46f4-87af-0213945876b8 (1)

No idea though is it related to migration problem or not.

Where else can I look into to find out why it is failing?

This is on 3 node cluster and the nodes are identical, except one node has 3x1Tb HDD while two others have 2x1Tb HDDs.
Userlevel 7
Badge +25
Yeah in virsh you could likely set the lowest common denominator if you could update cpu.model and have it stick. I think EVC basically does the same by masking the upstream features in the ISA.

I know a vm.flags override is on my wish list as well.

Have you tried to enable the ignore_msrs flag for KVM (talked in other areas to get W10 compat) to see if that might allow free movement? The KVM flag won't get wiped like the VM flags do.
Badge +5
I would just like to say that without an equivalent to VMware's EVC mode I can't use the CE edition. Huge bummer, but glad I found this post. I was planning on migrating my VMware enviroment to a 4 node CE cluster this weekend.

I have 3 different Intel CPUs. 1 Sandy Bridge, 1 Ivy Bridge, and 2 Haswell.

Will that cause migration failures? Im not familiar with KVM.
Badge +6
Hi, Tony


ha.getfailover_enabled: truenum_reserved_hosts: 0logical_timestamp: 1

nothing about "reserve_entire_host=true"

Thank you,
Jsiranee
Userlevel 2
Badge +13
That is because of the HA setting. If you reserve a host it will not allow you run VM's on it.

If you checkout the seeing
acli
reserve_entire_host=true

Tony
Userlevel 7
Badge +24
We don't currently support migration across different CPU types. This will be addressed in a future release. Also note however that while we will endeavor to continue supporting as many older CPU types as possible, they will fall off in terms of capabilities. It looks like we need some improvements in terms of error messages, but at some point if you have older and non-uniform CPU types in a CE cluster, we may be unable to do migrations.
Badge +6
Hi,JonI've got problem about vm migrationmy nutanix cluster composes of1.NTNX-2aa1baca-A Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz (192.168.10.61)2.NTNX-97d7c49d-A Intel(R) Core(TM) i7 CPU 960 @ 3.20GHz (192.168.10.63)3.NTNX-26349e86-A Intel(R) Core(TM) i5-4460 CPU @ 3.20GHz (192.168.10.67)I tested migrate centos6 vm with the following testcases with GUI1. start vm on NTNX-2aa1baca-A2. migrate from NTNX-2aa1baca-A to NTNX-97d7c49d-A result success3. migrate from NTNX-97d7c49d-A to NTNX-2aa1baca-A result success4. migrate from NTNX-2aa1baca-A to NTNX-26349e86-A result success5. migrate from NTNX-26349e86-A to NTNX-2aa1baca-A result error as followingOperation failed: Error while migrating vm 07428a0c-a444-4617-9105-eeac1daf2fcc: unsupported configuration: guest and host CPU are not compatible: Host CPU does not provide required features: invpcid, erms, bmi2, smep, avx2, bmi1, fsgsbase, abm, pdpe1gb, rdrand, f16c, avx, osxsave, xsave, aes, tsc-deadline, movbe, x2apic, pcid, fma, pclmuldq6. migrate from NTNX-26349e86-A to NTNX-97d7c49d-A result error as followingOperation failed: Error while migrating vm 07428a0c-a444-4617-9105-eeac1daf2fcc: unsupported configuration: guest and host CPU are not compatible: Host CPU does not provide required features: invpcid, erms, bmi2, smep, avx2, bmi1, fsgsbase, abm, pdpe1gb, rdrand, f16c, avx, osxsave, xsave, aes, tsc-deadline, movbe, x2apic, pcid, fma, pclmuldqIs this the limitaion of the migration or bug of acropolis?

Thank you,
Jsiranee.
Userlevel 6
Badge +29
~/data/acropolis.logs (or something along those lines, can't remember if it is data/logs or just /logs of the top of my head), on each CVM, should hold some good data.
Userlevel 1
Badge +10
Took the lazy way out, disabled and re-enabled HA clusterwide and it started working again. I'm curious if there's logs somewhere to try to trace what the actual issue was?