Hello,
I test ceilometer agent notification workload partition. I found it is too fragile.
The load is 1k cirrors vm. I make processing queue to 4 and workers to 1.
I can sure the network is ok. But in the ceilometer agent notification log i see this
http://paste.openstack.org/show/624795/
I am sure rabbitmq resource is enough. the status is here
http://paste.openstack.org/show/624796/
And after some dig in i found this.
https://stackoverflow.com/questions/383738/104-connection-reset-by-peer-socket-error-or-when-does-closing-a-socket-resu
Can someone join me to test it? Or is there someone know this why and how to fix this?
Thanks a lot a lot.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20171027/87292f35/attachment.html>