[ClusterLabs] Antw: Re: Antw: Delayed first monitoring

[Jan Pokorný]

On 13/08/15 10:38 +0200, Ulrich Windl wrote:
>>>> Miloš Kozák <milos.kozak at lejmr.com> schrieb am 13.08.2015 um 09:56 in
>>> Nachricht <55CC4DAA.4020206 at lejmr.com>:
> 
>> 
>> Dne 13.8.2015 v 09:26 Andrei Borzenkov napsal(a):
>>> On Thu, Aug 13, 2015 at 10:01 AM, Miloš Kozák <milos.kozak at lejmr.com>
>>> wrote:
>>>> However, this does not make sense at all. Presumably, the
>>>> pacemaker should get along with lsb scripts which comes from
>>>> system repository, right?
>>>> 
>>> Let's forget about pacemaker for a moment. You have system startup
>>> where service B needs service A. initscript for service A completes
>>> and script for service B is started but service A is not yet ready to
>>> be used.
>>> 
>>> This is a bug in startup script. Irrespectively of whether you use it
>>> with pacemaker or not.
>> 
>> I am sorry, but I didnt get the point..
>> 
>> If service A is not ready then service B should not be started. 
> 
> As you seem to be ignorant for advice:
> Yes, you are right: Service B should check whether service A is up before
> starzing itself.
> The easy change for the start script of B is to find aout what command was run
> before it to check whether the command before did everything OK by checking
> again itself.
> 
> [...]

The harder task for the sketched, relaxed (not strictly serialized, at
least per prerequisite-ordering) environment is for service B aware of
its prerequisite-ordered predecessor A to (also) decide if A is not by
any chance just proceeding with a startup sequence -- something
requiring a very detailed knowledge of its internals and being
prone to race-conditions anyway.

Hence reasonable, high-level, init systems require such startup
sequences to be completely finished by the time they acknowledge
service at hand as "started" and allow prerequisite-ordered successor
to join the game too.  Consequently, the responsibility for such
"is finished with startup (successfully or not)?" is deferred to the
lower-level dedicated startup recipes that should then signal this
back to the init system (e.g., by finishing only when the startup
is over) credibly to prevent mess ups.

Going full circle, if such assumption is broken in httpd initscript,
it should be fixed.

-- 
Jan (Poki)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20150813/fec30f5a/attachment-0003.sig>