Bug 881593
| Summary: | Incorrect VM count in the load column | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Virtualization Manager | Reporter: | Anush Shetty <ashetty> |
| Component: | ovirt-engine-webadmin-portal | Assignee: | Michal Skrivanek <michal.skrivanek> |
| Status: | CLOSED CURRENTRELEASE | QA Contact: | Pavel Stehlik <pstehlik> |
| Severity: | unspecified | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 3.1.0 | CC: | acathrow, bdagan, ecohen, grajaiya, iheim, jkt, michal.skrivanek, ofrenkel, Rhev-m-bugs, sdharane, vbellur, yeylon |
| Target Milestone: | --- | ||
| Target Release: | 3.3.0 | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Whiteboard: | virt | ||
| Fixed In Version: | is1 | Doc Type: | Bug Fix |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | Type: | Bug | |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
| Bug Depends On: | |||
| Bug Blocks: | 970195 | ||
|
Description
Anush Shetty
2012-11-29 07:34:51 UTC
The attached log is full of messages of type: 2012-11-28 23:37:13,844 INFO [org.ovirt.engine.core.vdsbroker.VdsUpdateRunTimeInfo] (QuartzScheduler_Worker-59) [5d3fbefb] VDS::UpdateVmRunTimeInfo Error: found VM on a VDS that is not in the database! So, probably up-to-date information cannot be fetched and updated. Could the both nodes be in inconsistent state? E.g. installation of VDSM of the first node run with success: ... 2012-11-28 23:00:43,385 INFO [org.ovirt.engine.core.bll.VdsInstaller] (NioProcessor-33) Installation of 10.70.36.1. Received message: <BSTRAP component='VDS Configuration' status='OK'/>. FYI. (Stage: Running second installation script on Host) ... The was also the another installation for the second host, but with some errors: ... 2012-11-28 23:22:48,175 INFO [org.ovirt.engine.core.bll.VdsInstaller] (NioProcessor-38) Installation of 10.70.36.6. Received message: <BSTRAP component='RHN_REGISTRATION' status='FAIL' message='Unable to fetch vdsm package. Please check if host is registered to RHN, Satellite or other yum repository'/>. FYI. (Stage: Running first installation script on Host) ... 2012-11-28 23:22:48,212 ERROR [org.ovirt.engine.core.bll.VdsInstaller] (pool-4-thread-47) [2e25840] Installation of 10.70.36.6 has failed. Failure details: SSH command failed while executing at host '10.70.36.6', refer to logs for further information. (Stage: Running first installation script on Host) ... From this time just UpdateVmRunTimeInfo error messages follow. I tried this with the latest ovirt-engine, 2 hosts with Fedora17 and stable VDSM package installed and running - I think, the behaviour is correct. Before I continue, please answer the questions... Do you think, the second host could be in incorrect state? Could the message 'Unable to fetch vdsm package. Please check if host is registered to RHN, Satellite or other yum repository' be an issue? What was the status of VDSM package on both hosts? Just clean machine, or some installation has happened before? Can this bug be reproduced - a) clean hosts, b) hosts with performed correct setup? Thank you. I'm pretty sure the load is incorrect when the VMs are in 'migrating' state. (In reply to comment #8) > I'm pretty sure the load is incorrect when the VMs are in 'migrating' state. Indeed, while migration in process the VM is counted twice - one per host, since you have a VM process on both hosts. I was mislead by the migration description since my first understanding of the description, it looked like the complaint is after the migration is over. And according to the message: 2012-11-28 23:37:13,844 INFO [org.ovirt.engine.core.vdsbroker.VdsUpdateRunTimeInfo] (QuartzScheduler_Worker-59) [5d3fbefb] VDS::UpdateVmRunTimeInfo Error: found VM on a VDS that is not in the database! It looks like there is independent VM on the host. On the host load the engine reports all VMs VDSM reports, including VMs that where not started by the engine. Possible reason for the floating VM in a migration scenario: Migration finished successfully but the original host thinks migration failed and does not kill the original machine. But!!!! Reading again the description the extra VM appeared in the system before the migration, from comment 0: 1. add rhev-h 2. create vm 3. add another rhev-h to the same cluster 4. load column shows 1 VMs each. So I wander where this extra VM came. According to the log the engine it did find an extra unexpected VM on the stem. Anush, could it be that you've missed a step? Was the VM an HA VM? Did you had trouble with the first host prior to adding the second? (In reply to comment #9) > (In reply to comment #8) > > I'm pretty sure the load is incorrect when the VMs are in 'migrating' state. > > Indeed, while migration in process the VM is counted twice - one per host, > since you have a VM process on both hosts. Yet it is seen by none of them. I just migrated 15 VMs from host A to B - they both showed 0 VMs while migrating, and when all was done, it got back to normal. My observation is the following: During migration of *one* VM the vm_count=1 for *both* host. This is correct behaviour, the VM runs on both hosts. After migration ends, the vm_count=1 is on just one host. vd.vm_active=1 just for one host always. The GUI table is filled using e.g. this query: select vs.vds_name,vd.vm_count, vd.vm_active from vds_static vs join vds_dynamic vd on vd.vds_id = vs.vds_id I do not understand: ...they both showed 0 VMs while migrating... Can you try to run the query on your configuration during migration? Can we close this issue? Here's an example: I have two hosts: Host A: 16 VMs. Host B: 10 Vms. I migrate 4 from Host A to Host B, then switch to a different tab and go back to the hosts tab. Now I see: Host A: 12 VMs. Host B: 10 VMs. Again: Host A: 12 VMs. Host B: 14 VMs. There is no VM, at any point, in the host's 'Virtual Machines' subtab that shows ' migrating from' status. Current behaviour for RHEV 3.2 (sf6) is the following (target release): Host A: 1 VM Host B: 0 VM During migration: Host A: 1 VM Host B: 1 VM After migration: Host A: 0 VM Host B: 1 VM In the version 3.1 it looked differently. Please, confirm the correct behaviour and tested version. Current behavior for RHEV 3.1 (si26.1), load column in "Hosts" tab is the following: Host A: 5 VM Host B: 0 VM During migration: Host A: 0 VM Host B: 0 VM After migration: Host A: 0 VM Host B: 5 VM The behavior for RHEV 3.2 (sf6) is different indeed, (Virtual Machines column, "Hosts" tab): Host A: 1 VM Host B: 0 VM During migration: Host A: 1 VM Host B: 1 VM After migration: Host A: 0 VM Host B: 1 VM Note, The hosts are running RHEL 6.3, 6.4 respectively VM count during migration for 3.2 looks without problems. The VM on a target host is taking resources, so it is useful/important to have it in the statistics. We are investigating an option to show the VM on target host in 'Migrating To' status (currently the VM is shown just on the source host Virtual Machine subtab in Migrating From status) during the migration. Merged u/s: 9ba2d45fca8726279b9bace2d2fae7beff110e4f *** Bug 635980 has been marked as a duplicate of this bug. *** Closing - RHEV 3.3 Released Closing - RHEV 3.3 Released Closing - RHEV 3.3 Released |