Hi, Jiaju I made the daemon who supervises the resource depending on a ticket, in order to solve this problem.
I have sent the following "pull request". https://github.com/jjzhang/booth/pull/52 The feature is as follows. - The information on the ticket to supervise is acquired from the configuration file of booth. - A ticket becomes "grant", and if a resource start(s), surveillance will start. - booth_resource_monitord moves a ticket to other sites using booth, when it becomes impossible for a resource to work in a site. - booth_resource_monitord will be installed when the configure option was with the "--enable-resource-monitor". How to use: Usually, booth_resource_monitord is added to the composition which is using booth as follows. =================================================================== group grpBooth prmIpBooth prmApBooth prmApBooth_rsc_mond primitive prmIpBooth ocf:heartbeat:IPaddr2 \ params ip="***.***.***.***" nic="eth*" cidr_netmask="24" \ op start interval="0s" timeout="60s" on-fail="restart" \ op monitor interval="10s" timeout="60s" on-fail="restart" \ op stop interval="0s" timeout="60s" on-fail="fence" primitive prmApBooth ocf:pacemaker:booth-site \ op start interval="0s" timeout="90s" on-fail="restart" \ op monitor interval="10s" timeout="60s" on-fail="restart" \ op stop interval="0s" timeout="100s" on-fail="fence" primitive prmApBooth_rsc_mond ocf:heartbeat:anything \ params binfile="booth_resource_monitord" \ op start interval="0s" timeout="90s" on-fail="restart" \ op monitor interval="10s" timeout="60s" on-fail="restart" \ op stop interval="0s" timeout="100s" on-fail="fence" -------------------------------------------------------------------- limitation: The target resource cannot be read when "rsc_ticket" is described by "resource_set". I want me to merge this function into the sauce tree of booth by all means. Best Regards, Yusuke (2012/03/08 11:37), Yuusuke Iida wrote: > Hi, Jiaju > > Thank you for reply. > > (2012/03/05 14:00), Jiaju Zhang wrote: >> Hi Yuusuke, >> >> On Mon, 2012-03-05 at 11:49 +0900, Yuusuke Iida wrote: >>> Hi, Jiaju >>> >>> I thought about a plan to deal when a resource did not change in sites. >>> I think that I make daemon working outside booth. >>> >>> This daemon watches it whether a resource can work in sites. >>> And it executes revoke command for booth when the state that a resource >>> cannot manage was confirmed. >>> booth catches revoke and thinks that I move a ticket to another site. >> >> If I understand it correctly, the daemon you mentioned automated some of >> the admin's behaviors, if the resources cannot be managed by one site, >> revoke the ticket and move the ticket to another site. I have no >> objection if the admin has this requirement;) > Thank you for agreeing. > The summary of the processing is just what you think. > admin may not necessarily need this function. > However, I think that admin which wants to automate processing as much > as possible exists. > >> The only thing I'm not sure is if the admin really want to do this? My >> assumption is if the local site is alive the admin will be inclined to >> keep the ticket stay in this site, if the site is totally down, we have >> no choice, the ticket has to move to another site to keep the service >> available. >> However, that is just one using scenario in my mind, booth should >> support the using scenario that you mentioned;) >> >>> >>> I think that the continuity of the resource is kept in this movement. >>> >>> I analyze CIB and intend to perform the state confirmation of the >>> resource using score. >> >> I'm not quite understand here, do you mean that if the resource usually >> being un-managed by this site, we'd better move it to another site, so >> your daemon will depends on this value to decide whether it would move >> the ticket another site, right? > When a resource failed, I think that the score of the resource becomes > less than 0. > When the resource was not able to start in all nodes in the site, I > think that score becomes less than 0 in all nodes. > I want to judge the state that a resource was not able to operate from > this score. > > When a ticket does not become grant, the score of the resource becomes > less than 0. > Therefore, I want to monitor the resource while a ticket becomes grant. > >> >> Well, I think you raised another using scenario which I has not thought >> of before;) And I agree with you to setup such a daemon to do this work >> if the admin need. > I want you to confirm it again when you were completed. > > Thanks, > Yuusuke >> >> Thanks, >> Jiaju >> >> > -- ---------------------------------------- METRO SYSTEMS CO., LTD Yuusuke Iida Mail: iiday...@intellilink.co.jp ---------------------------------------- _______________________________________________ Pacemaker mailing list: Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
