Previously, the host's committed memory was not cleared immediately after power outage.
This meant that it was not possible to run VMs on the host immediately after power outages if the committed memory of the host was high when it was up. If there were highly-available VMs running on the host and there is no other active host in the cluster, failure to restart them meant that they were not restarted automatically.
Now, committed memory of hosts is calculated before we try to restart highly available VMs that were running on it before it crashed.
When the only active host in a cluster crashes due to power outage, highly-available VMs running on it are now automatically restarted when the host is rebooted.
Created attachment 876652[details]
engine.log
Description of problem:
I have one host with 16Gb and I run on it 4 HA vms 3 with 4096Mb and 1 with 2048Mb vms started and runs fine.
After it I poweroff host, wait 5 minutes and power on host, vms failed to start first time:
CanDoAction of action RunVm failed. Reasons:VAR__ACTION__RUN,VAR__TYPE__VM,ACTION_TYPE_FAILED_VDS_VM_MEMORY,SCHEDULING_ALL_HOSTS_FILTERED_OUT,VAR__FILTERTYPE__INTERNAL,$hostName aqua-vds2.qa.lab.tlv.redhat.com,$filterName Memory,SCHEDULING_HOST_FILTERED_REASON
After it engine try to start HA vms second time and success.
Version-Release number of selected component (if applicable):
av3
How reproducible:
Always
Steps to Reproduce:
1. See above
2.
3.
Actual results:
Engine failed to start HA vms first time because memory filter
Expected results:
Engine success to start HA vms first time, without any error in engine log
Additional info:
Bug also exist in 3.3(checked for is35) and it more critical, because after first failed, engine not try to start HA vms second time
Created attachment 876652 [details] engine.log Description of problem: I have one host with 16Gb and I run on it 4 HA vms 3 with 4096Mb and 1 with 2048Mb vms started and runs fine. After it I poweroff host, wait 5 minutes and power on host, vms failed to start first time: CanDoAction of action RunVm failed. Reasons:VAR__ACTION__RUN,VAR__TYPE__VM,ACTION_TYPE_FAILED_VDS_VM_MEMORY,SCHEDULING_ALL_HOSTS_FILTERED_OUT,VAR__FILTERTYPE__INTERNAL,$hostName aqua-vds2.qa.lab.tlv.redhat.com,$filterName Memory,SCHEDULING_HOST_FILTERED_REASON After it engine try to start HA vms second time and success. Version-Release number of selected component (if applicable): av3 How reproducible: Always Steps to Reproduce: 1. See above 2. 3. Actual results: Engine failed to start HA vms first time because memory filter Expected results: Engine success to start HA vms first time, without any error in engine log Additional info: Bug also exist in 3.3(checked for is35) and it more critical, because after first failed, engine not try to start HA vms second time