CeTune

CeTune copied to clipboard

Hello guys , I am running CeTune on CentOS nodes and got stuck at haproxy step. On RHEL based machines there is no /etc/default/haproxy configuration file so CeTune fails , is this a known issue ?

BTW haproxy rpm is properly installed on these nodes

[2016-05-08T11:51:07.632079][LOG]: Ceph already installed on below nodes
[2016-05-08T11:51:07.632285][LOG]: Install opts is '--release '
[2016-05-08T11:51:08.581212][WARNING]: Found different configuration from conf/ceph_current_conf with your desired config : OrderedDict([('mon', {}), ('osd', {}), ('mds', {}), ('osd_num', 3), ('radosgw', [])])
[2016-05-08T11:51:08.582611][LOG]: Generating rgw ceph.conf parameters
[2016-05-08T11:51:08.582849][LOG]: configure node2-1 in ceph.conf
[2016-05-08T11:51:13.282435][LOG]: deploy radosgw instances
[2016-05-08T11:51:19.644885][LOG]: Creating rgw required pools
[2016-05-08T11:58:03.390218][LOG]: Updating haproxy configuration
[2016-05-08T11:58:05.927822][ERROR]: node2: sed: can't read /etc/default/haproxy: No such file or directory
pdsh@node1: node2: ssh exited with exit code 2

As a temporary fix i created a dummy file /etc/default/haproxyand it worked

In addition to that i am also seeing problems with haproxy service restart by CeTune

[2016-05-08T15:31:04.080848][LOG]: Generating rgw ceph.conf parameters
[2016-05-08T15:31:04.081019][LOG]: Distribute ceph.conf
[2016-05-08T15:31:08.846870][LOG]: Restart ceph cluster
[2016-05-08T15:31:08.855572][LOG]: Shutting down mon daemon
[2016-05-08T15:31:09.277531][LOG]: Shutting down osd daemon
[2016-05-08T15:31:10.246832][LOG]: Starting mon daemon
[2016-05-08T15:31:11.014001][LOG]: Started mon.node1 daemon on node1
[2016-05-08T15:31:11.014233][LOG]: Starting osd daemon
[2016-05-08T15:31:11.524375][LOG]: Started osd.0 daemon on node1
[2016-05-08T15:31:12.154690][LOG]: Started osd.1 daemon on node2
[2016-05-08T15:31:12.605467][LOG]: Started osd.2 daemon on node3
[2016-05-08T15:31:12.612739][LOG]: Restart radosgw
[2016-05-08T15:31:14.177886][ERROR]: node2: bash: /etc/init.d/haproxy: No such file or directory
pdsh@node1: node2: ssh exited with exit code 127

hi, @ksingh7 , deploy rgw part is quite hard coded, to make sure everything goes well after running 'deploy rgw', we add haproxy configuration and restart, but which only works on debian(apt-get installed ).I am thinking the more reasonable way is to remove 'haproxy deployment codes' away from 'deploy rgw codes', do you think that will be more workable in your environment?

Hi @xuechendi In my opinion haproxy is useful in case we want to run COSbench over multiple RGW instances to test scalability and performance out of multiple RGW instances.

I guess as a long term solution we need to make haproxy part more flexible with a check like this

if os==ubuntu/debian 
 do this 
else if os==rhel 
 do this 

I am not sure how difficult it is to implement this ? Do you have any alternate thoughts to fix this ?

I didn't go that way is because I don't have a rhel setup myself, but I think to separate the handling by os is the first thing I need to add.

It that ok for you to share a centos haproxy default and rgw deployed configuration to me,so I can just fix this on my Ubuntu setup?

发自我的 iPhone 在 2016年5月9日，18:18，karan singh <[email protected]mailto:[email protected]> 写道：

Hi @xuechendihttps://github.com/xuechendi In my opinion haproxy is useful in case we want to run COSbench over multiple RGW instances to test scalability and performance out of multiple RGW instances.

I guess as a long term solution we need to make haproxy part more flexible with a check like this

if os==ubuntu/debian do this else if os==rhel do this

I am not sure how difficult it is to implement this ? Do you have any alternate thoughts to fix this ?

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHubhttps://github.com/01org/CeTune/issues/64#issuecomment-217827722

Unfortunately i don't have working haproxy config file for CentOS system , i hope if you have a Ubuntu one then tailoring it for CentOS should not be much change ?