Hi Steve,

On Mon, 2008-06-23 at 21:22 -0700, Steven Dake wrote:
> Shridhar,
> 
> Yes there has been a forward port of the sync and totem code from
> whitetank to address these issues mailed to this list already.  This,
> along with other improvements, is going to be released today as
> openais 0.84 (unstable).

Yes I had seen those mails. But hasnt it been committed to the svn yet?
Because I tried using the latest sync and encountered same problem.


> Regards
> -steve
> 
> On Mon, Jun 23, 2008 at 5:43 PM, Shridhar Sahukar
> <[EMAIL PROTECTED]> wrote:
>         Hi Steve,
>         
>         I was running openais-0.82 on solaris and found that there is
>         some major
>         issue with the totem code which has already been fixed in the
>         whitetank.
>         I took the latest of sync from trunk but that seems to have
>         same
>         problem. Finally I took totemsrp.c from the whitetank and the
>         issue
>         seems to have been fixed here.
>         
>         Below is the details of the problem:
>         
>         I brought up openais on 2 solaris machines. Both of them join
>         the ring
>         properly. But when a node is leaving, the my_member_entries in
>         the srp
>         code seems have been removed before the config change callback
>         completes.
>         For ex, in your clm code, when you get a config change
>         callback, you map
>         the nodeId of the leaving node to the corresponding IP address
>         by
>         invoking totempg_ifaces_print(). This inturn invokes
>         totemsrp_ifaces_get() where you search for the given nodeId in
>         the
>         instance->my_member_list.
>         
>         In the trunk code I am observing that the nodeId is already
>         being
>         removed from the member_list, and hence in the config_ch
>         function you
>         wont get the IP address of the leaving node.
>         
>         I remember that we had found a similar issue when we were
>         debugging the
>         mixed endian problems, where the memebr_list was getting
>         updated before
>         the comit token or something of that sort.
>         
>         
>         Could you please look into this?
>         
>         Also, as I had discussed with you some time back, the trunk
>         code doesnt
>         compile properly on Solaris. I did some custom modifications
>         to the make
>         files to suite our requirement. I can send you the diff of it.
>         Please
>         let me know if can spare time to look into this issue.
>         
>         Regards,
>         Shridhar
>         
>         
>         _______________________________________________
>         Openais mailing list
>         [email protected]
>         https://lists.linux-foundation.org/mailman/listinfo/openais
> 

_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais

Reply via email to