On 2008-10-13T13:38:44, Florian Haas <[EMAIL PROTECTED]> wrote: > Hello, > > inspired by a discussion with the SerNet guys at Linux Kongress last > week, here's a thought I'd like to poll comments on. > > Clustered Samba (with ctdb) currently uses its own cluster manager. Part > of the reason for this seems to be that none of the available cluster > managers (Heartbeat and RHCS at the time of implementation) had support > for what the Samba guys call "tickle ACKs". See > http://tinyurl.com/4kul8z for an article discussing this.
Right, that's a very interesting feature which got implemented in CTDB there. Thankfully to Open Source, instead of going their own route, they were able to chose to extend an existing solution instead of reinventing the wheel - oh, sorry, I got caught in a dreamworld here ;-) > Now I wonder if one could add functionality to the IPaddr2 RA to achieve > in essence the same thing. Suppose that triggered by an optional > resource parameter, IPaddr2 would invoke a mechanism similar to the one > employed by cutter (http://www.lowth.com/cutter/) after IP address > takeover. In conjunction with a TCP connection state replication utility > such as conntrackd (http://conntrack-tools.netfilter.org/), this should > enable the RA to actively cut off TCP connections to that IP address, > forcing a client reconnect. Cutter itself will only work on forwarded > connections, so unless one wants to change cutter, I wonder whether a > different tool would be of better use. > > I haven't tested this much myself, but wanted to see if there are any > drawbacks or obvious dead ends here that I didn't think of before I > start to do so. So I'm thankful for any and all comments. I very much agree, and even added this to our todo list. I would of course welcome if someone else does this ;-) Some comments: - I believe IPaddr2 is the right place to add this. This should be a feature generically available to all IP addresses and TCP services being managed, and I would argue that a flag for the IPaddr2 RA is the best place for this. The RA would need to start an additional daemon for this, and IPaddr2 actually already has some clone extensions. - Replication conntrackd is a great base to start with I think, but I'd really prefer if the sync'ing went through openAIS too, instead of needing to be actively configured. - cutter I think cutter might even work, as the replicated nodes do think that they are "forwarding" the connection. We just need to identify all connections which passed through the failed node. (Remember that for Cluster Alias IP, that might not be all flows associated with the IP.) Regards, Lars -- Teamlead Kernel, SuSE Labs, Research and Development SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) "Experience is the name everyone gives to their mistakes." -- Oscar Wilde _______________________________________________________ Linux-HA-Dev: [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
