Bug 984287

David,

As I understand from Yaniv.D this is different only in the stop service not in the satart, please confirm.

In addition I would like to know (on the same host) how much it takes for the stop (and restart) opertions to happen without having the Reports packages  installed.

Hi,

Please describe the matrix of testing in order to avoid misunderstanding:
Suggestion:

On a bare metal server
-----------------------
1. rhevm3.2 without reports
stop ovirt-engine = time ?
restart ovirt-engine = time ? 

2. rhevm3.2 with reports
stop ovirt-engine = time ?
restart ovirt-engine = time ?
connecting time to reports portal ?

3. rhevm3.3 without reports
stop ovirt-engine = time ?
restart ovirt-engine = time ?

4. rhevm3.3 with reports
stop ovirt-engine = time ?
restart ovirt-engine = time ?
connecting time to reports portal ?

Today we performed some tests with two identical machines (in terms of hardware power) and these are the results:

1. Machine with RHEV-M 3.2 and reports.

This machine needs 37s since issuing "service ovirt-engine restart" till the reports login page is available in the browser. Of those 37 seconds 13 correspond to the deployment of the application, as can be seen in server.log:

2013-07-22 10:29:37,791 INFO  [org.jboss.as.server.deployment] (MSC service thread 1-3) JBAS015876: Starting deployment of "rhevm-reports.war"
2013-07-22 10:29:50,643 INFO  [org.jboss.as.server] (ServerService Thread Pool -- 17) JBAS018559: Deployed "rhevm-reports.war"

In this machine the service (including reports) takes 1.2 seconds to restart:

[root@aqua-vds1 ~]# time service ovirt-engine restart
Stopping engine-service: [  OK  ]
Starting engine-service: [  OK  ]

real	0m1.263s
user	0m0.142s
sys	0m0.094s

And approx the same to stop:

[root@aqua-vds1 ~]# time service ovirt-engine stop
Stopping engine-service: [  OK  ]

real	0m1.139s
user	0m0.087s
sys	0m0.048s

2. Machine with RHEV-M 3.3 and reports.

This machine needs 43s since issuing "service ovirt-engine restart" till the reports login page is available in the browser. Of those 43 seconds 15 correspond to the deployment of the application, as can be seen in server.log:

2013-07-22 10:24:35,071 INFO  [org.jboss.as.server.deployment] (MSC service thread 1-9) JBAS015876: Starting deployment of "rhevm-reports.war"
2013-07-22 10:24:49,702 INFO  [org.jboss.as.server] (ServerService Thread Pool -- 19) JBAS018559: Deployed "rhevm-reports.war"

In this machine the service (including reports) takes 1.3s to restart:

[root@aqua-vds8 ovirt-engine]# time service ovirt-engine restart
Stopping oVirt Engine:                                     [  OK  ]
Starting oVirt Engine:                                     [  OK  ]

real	0m1.318s
user	0m0.142s
sys	0m0.065s

And approx the same to stop:

[root@aqua-vds8 ovirt-engine]# time service ovirt-engine stop
Stopping oVirt Engine:                                     [  OK  ]

real	0m1.133s
user	0m0.018s
sys	0m0.006s

So the 3.3 machine is a bit slower than the 3.2 machine, approx 6 seconds slower. Taking into account that this is a manual process and that I am connecting remotely I think that this difference isn't meaningful. Also, even I tested repeatedly, I didn't find any situation where the engine takes more than 2 seconds to stop. However I remember seeing that situation during the previous tests that David and I did together, so I think is worth trying to find what triggers it. These two environments are completely empty, no hosts or VMs, maybe it is worth testing with environments that are populated with VMs as it may happen that some of the threads that we start refuse to stop if busy, thus delaying the stop operation.

All in all, if we are able to reproduce the long stop time I can continue studying it, otherwise I can't make any progress.