[Pacemaker] [Linux-HA]I revised resource-agents of tomcat.
nozawat
nozawat at gmail.com
Thu Feb 9 11:57:50 UTC 2012
Hi Brett / Dejan
>The string search is based upon the premise that every tomcat resource
>agent must have a unique name.
I think so, too.
>If this name is used in the CATALINA_OPTIONS variable it will be
>visible in the process table.
>To be honest, I've never liked this and actually prefer the process ID check.
I thought that it was enough only in -Dname if a process check was possible.
However, I judged that it was necessary for the process check method
because Brett added CATALINA_OPTS in old days.
I may not move when I add CATALINA_OPTS by a process check method.
>I believe it was originally done to support ease of resource agent
>configuration as at the time, I don't think the
>process ID operations using Tomcat's start/stop scripts were
>particularly reliable.
Handling of PID in the Tomcat script seems to be surely a transition period.
However, I believe that I can cope with the following patches.
<https://github.com/ClusterLabs/resource-agents/commit/7d9e8a6250ea084f77547dff3bea79974987a708#heartbeat/tomcat>
Regards,
Tomo
2012年2月9日18:31 Brett Delle Grazie <brett.dellegrazie at gmail.com>:
> Hi Dejan / Tomo,
>
> Apologies for the long delay in response but I've been scratching my
> head as to some of the patches
> caused me some concern.
>
> Specifically:
> 1) Patch #2 from the first link
> (http://www.gossamer-threads.com/lists/linuxha/dev/77116)
> This changes the 'status' check to a 'monitor' operation.
>
> The monitor operation should be configured by the user to point at the
> application running within tomcat, e.g. a HEAD request to a version
> page or similar.
>
> However this operation can and will frequently block for an
> indeterminate period of time.
>
> It might be worthwhile making the 'status' operation a process-id
> check for _tomcat_ rather than a monitor operation which traditionally
> checks the availability of the application running within tomcat.
>
> There are distinct conditions where Tomcat may still be working but
> the applications running on it are dead in the water.
>
> This all depends upon whether there is value in distinguishing between
> 'tomcat is dead' and 'the app running on tomcat is dead but tomcat is
> fine'.
>
> 2) The patch contained in the second link and possibly some of the
> patches in the first link.
> The string search is based upon the premise that every tomcat resource
> agent must have a unique name.
> If this name is used in the CATALINA_OPTIONS variable it will be
> visible in the process table.
> To be honest, I've never liked this and actually prefer the process ID check.
> I believe it was originally done to support ease of resource agent
> configuration as at the time, I don't think the
> process ID operations using Tomcat's start/stop scripts were
> particularly reliable.
>
> Tomcat's come a long way in that time and perhaps we could use both. i.e.
> If the PID is defined, use it, if its not defined, use the string search.
> In either case - I think the -Dname option is to ambiguous (what
> happens if the application requires a 'name' parameter to be set on
> the command line?) we should probably increase this to
> -Dtomcat_ra_name=<name> or something similar.
>
> These are just observations and may no longer be correct as its been
> about a year since I've used tomcat resource agent.
>
> I'm not actively monitoring the list right now so please cc me if you
> wish me to comment further.
>
> Thanks
>
> Kind regards,
>
> Brett
>
> On 23 January 2012 20:23, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>> Hi,
>>
>> On Thu, Jan 19, 2012 at 02:00:31PM +0900, nozawat wrote:
>>> Hi Brett
>>
>> Adding Brett to the list of recipients :)
>>
>> Cheers,
>>
>> Dejan
>>
>> P.S. Always good to just send again to the ML too.
>>
>>> I want comment if possible about a patch of following tomcat.
>>>
>>> <http://www.gossamer-threads.com/lists/linuxha/dev/77116>
>>> ->I did three trifling corrections.
>>>
>>> <http://www.gossamer-threads.com/lists/linuxha/dev/75950#75950>
>>> ->I change the check method of the process.
>>>
>>> Regards,
>>> Tomo
More information about the Pacemaker
mailing list