[Linux-HA] hb master proc leaking memory on client signon/off?

Guochun Shi gshi at ncsa.uiuc.edu
Tue Mar 22 16:04:27 MST 2005


I see. I will look into it. Thanks for reporting the problem.

-Guochun

At 05:41 PM 3/22/2005 -0500, forrest whitcher wrote:
>Guochun,
>
>Yes, I've done that (it's how/why I noticed it in the first place) and it
>does continue increasing the heartbeat process's mem 4k each 
>invocation. I have run for 100's of invocations, and it just keeps
>adding ~4k.
>
>fw
>
>On Tue, 22 Mar 2005 16:18:16 -0600, Guochun Shi <gshi at ncsa.uiuc.edu> wrote:
>> one test may not be enough:
>> http://www.ussg.iu.edu/hypermail/linux/kernel/0207.1/1022.html
>> If  you test it lots of times, say 100, and it keeps increasing each time,
>> then we may have problem.
>> 
>> BTW 4k is the bufpool size in IPC.
>> 
>> -Guochun
>> 
>> At 04:49 PM 3/22/2005 -0500, you wrote:
>> >Alan,
>> >
>> >By reproduce, do you mean does running cl_status increase the memory
>> >in use b y the heartbeat master daemon?
>> >
>> >If that's what you mean then yes it does, same thing, about 4k each time
>> >I run cl_status.
>> >
>> >fw
>> >
>> >
>> >On Mon, 21 Mar 2005 17:07:03 -0700, Alan Robertson <alanr at unix.sh> wrote:
>> >> forrest whitcher wrote:
>> >> > Alan et al
>> >> >
>> >> > I've been kicking the tires some on Heartbeat and have observed that every
>> >> > time a new client connects (signon-s) to the heartbeat process the the master
>> >> > process ratchets up it's memory use by about 4K of ram.
>> >> >
>> >> > This is when I do the following:
>> >> >
>> >> > hb = ll_cluster_new("heartbeat");
>> >> >  signon(hb, MY_NAME);
>> >> >  cl_log_set_facility(facility);
>> >> >   init_nodewalk(hb);
>> >> >    nextnode(hb));          /* repeat while more nodes */
>> >> >   end_nodewalk(hb);
>> >> >  signoff(hb);
>> >> > delete(hb);
>> >> >
>> >> > I'm checking for HA_OK , so I know all the calls have completed, -- is there
>> >> > a reason for the daemon to hold the connection resources after exit??.
>> >> >
>> >> > I'm running 1.99.2 pulled from CVS about a week ago
>> >>
>> >> No.  This is a bug.  It's something we fixed in the past, but it seems to
>> >> have recurred.  Can you reproduce it using cl_status?
>> >>
>> >> --
>> >>      Alan Robertson <alanr at unix.sh>
>> >>
>> >> "Openness is the foundation and preservative of friendship...  Let me claim
>> >> from you at all times your undisguised opinions." - William Wilberforce
>> >>
>> >_______________________________________________
>> >Linux-HA mailing list
>> >Linux-HA at lists.linux-ha.org
>> >http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> 
>>




More information about the Linux-HA mailing list