Ask Your Question
0

RetryFilter: What to do with a failed host?

asked 2015-04-28 19:09:14 -0500

Bernd Bausch gravatar image

If I understand RetryFilter correctly, a host that once failed will be filtered out forever. Is that correct? If so, how can I convince the filter that the host is available again?

By the way, on my Juno installation on Centos 7 I find the following in the nova-scheduler.log:

Host [u'compute1', u'compute1'] fails.  Previously tried hosts: [[u'compute1', u'compute1']] host_passes /usr/lib/python2.7/site-packages/nova/scheduler/filters/retry_filter.py:42

It's a bit unclear to me what that means. First it says that compute1 fails, then it says that it passes. Which of the two?

edit retag flag offensive close merge delete
add a comment

2 answers

Sort by ยป oldest newest most voted
0

answered 2017-11-14 23:11:50 -0500

Sam Song gravatar image

The log doesn't mean compute1 has ever passed but just tried and failed. I have the same problem as yours. Do you find out how to convince the filter the the host is available again?

edit flag offensive delete link more

Comments

Today I know more than 2.5 years ago. It works as follows: compute1 was selected by scheduler, instance launch on compute1 failed, scheduler tries again but this time RetryFilter excludes compute1 to avoid endless loops.

At the next instance launch, compute1 is again included in the list.

Bernd Bausch gravatar imageBernd Bausch ( 2017-11-21 17:59:09 -0500 )edit
add a comment
0

answered 2015-04-29 00:54:44 -0500

dbaxps gravatar image

updated 2015-04-29 00:55:55 -0500

It might be Nova reaction on KVM/Libvirt failure to start instance on Compute Node.

1. Check `nova hypervisor-list`  contains compute node names 
2. Check openstcak-nova-compute is up and running on Compute nodes
3. Check instances logs under /var/log/libvirt/qemu ( or /var/log/libvirt/libxl case of Xen)

Via my experience with nova libvirt-xen driver real errors to troubleshoot came from libvirtd daemon on Compute node. Nova just reported "unable schedule instance"

edit flag offensive delete link more
add a comment

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

subscribe to rss feed

Stats

Asked: 2015-04-28 19:09:14 -0500

Seen: 1,756 times

Last updated: Apr 29 '15

Related questions

how to configure nova scheduler/filter and memory usage of compute nodes?

how security group is implemented

Got ERROR log when "nova-manage db sync"

gluster volume for nova instance

Launching instance on RHOSP 10 (Newton) fails with memory over 30G

ERROR: The server has either erred or is incapable of performing the requested operation. (HTTP 500)

nova-docker setup in devstack

installing nova from git

Where is the documentation for blueprint shelve-instance?

Ceilometer filter by user_id (or username)