Re: [Linux-ha-dev] Re: [Linux-HA] Recovering from unexpected bad things - is STONITH the answer?
Alan Robertson wrote: Kevin Tomlinson wrote: On Tue, 2007-11-06 at 10:25 -0700, Alan Robertson wrote: We now have the ComponentFail test in CTS. Thanks Lars for getting it going! And, in the process, it's showing up some kinds of problems that we hadn't been looking for before. A couple examples of such problems can be found here: http://old.linux-foundation.org/developer_bugzilla/show_bug.cgi?id=1762 http://old.linux-foundation.org/developer_bugzilla/show_bug.cgi?id=1732 The question that comes up is this: For problems that should never happen like death of one of our core/key processes, is an immediate reboot of the machine the right recovery technique? snip Here's the issue: The solution as I see it is to do one of: a) reboot the node and clear the problem with certainty I'm well aware that saying yeah, we did it that way isn't Good Form, but, well, we did it that way (also we did 'd', see below). I refer to existing proprietary HA products in which I was involved designing and implementing. What we found was that certain processes were indistinguishable from the node itself and their failure was therefore near impossible to deal with cleanly. The problem was, as described here, that the OS and applications/services would continue on but the other nodes in the cluster would see it as a node failure and take recovery actions (as you describe here). This was admin-controllable, as we did offer something else... b) continue on and risk damaging your disks. c) write some new code to recover from specific cases more gracefully and then test it thoroughly. d) Try and figure out how to propagate the failure to the top layer of the cluster, and hope you get the notice there soon enough so that it can freeze the cluster before the code reacts to the apparent failure and begins to try and recover from it. Our architecture was that our clients were in a 'process group' across the cluster (group services) where each was connected to our server process via a unix domain socket on the same node. Across the cluster these process group peers were assumed to be controlling resources and they mediated recovery actions through group services. The unix-domain socket breaking was (well, still is) a defined condition that the client was told to handle as our process dying. They were told in this case to immediately clean up and assume that their peers would assume that the local process was gone since the node had died and would be doing takeovers. This was meant to allow some hope of avoiding taking the node down for real, depending on the application space. The intent was that the local client would notice the death immediately, while the remote ones would take some time (i.e., lack of heartbeats, etc.) to notice. In cases where the node wasn't taken down, via inittab or similar our processes would get automatically restarted, and we'd reintegrate into the cluster. The local client process(es) would be expected to reconnect and have to rejoin their group(s). Our interface manual described all of this. If for some reason our processes couldn't restart (or inittab gave up because of too many retries) that node would stay out of the cluster. In the current code, sometimes you'll get behavior (a) and sometimes you'll get behavior (b) and sometimes you'll get behavior (c). In the particular case described by bug 1762, failure to reboot the node did indeed start the same resource twice. In a cluster where you have shared disk (like yours for example), that would probably trash the filesystem. Not a good plan unless you're tired of your current job ;-). I'd like to take most/all of the cases where you might get behavior (b) and cause them to use behavior (a). If writing correct code and testing it were free, then (c) would obviously be the right choice. Quite honestly, I don't know how to do (d) in a reliable way at all. It's much more difficult than it sounds. Among other reasons, it relies on the components you're telling to freeze things to work correctly. Since resource freezes happen at the top level of the system, and the top layers need all the layers under them to work correctly, getting this right seems to be the kind of approach you could make into your life's work - and still never get it right. You're right. In the scheme I described above we (group services) simply washed our hands of what our clients (layers above) were able to do and get right... We didn't write those. We offered this as the only thing we could think of, with the hope that some clients could do things correctly. It assumes that the OS, for example, is still working so the client can take dependable actions. And if they weren't confident they could enable node rebooting in this case and let recovery happen 'normally'. Case (c) has to be handled on a case by case basis, where
[Linux-ha-dev] Re: [Linux-HA] Recovering from unexpected bad things - is STONITH the answer?
Alan Robertson wrote: Yan Fitterer wrote: Not always. The case I have encountered (live) doesn't relate to HB component failure per se, but is nevertheless destructive. With an eDirectory load (and other database-backed software with large or lazily flushed write buffers would be similarly affected, IMHO), a hard reset of a node has a high likelihood of corrupting the database. This is in some cases no less destructive than allowing concurrent access to, say, an ext3 filesystem... If your software cannot withstand a crash, then it cannot be made highly-available - end of story. Crashes will happen. Be prepared. This is a fine argument from an engineering perspective, but not much use from a sysadmin POV. Heartbeat should (can and does!) help on any kind of software. I'm simply pointing out that (for less perfect software, amongst other reasons) the less STONITH (hard reset) potential cases we have, the better. :) Anything to avoid STONITH (in particular when a node isn't quite dead from the workload perspective). I have been pondering for a while the possibility of using some disk-based heartbeat to block STONITH, in cases where the STONITH target is still writing its disk heartbeat. This would in this case prevent data damage. In addition, I have been thinking of complementing this mechanism with a disk-based STONITH (otherwise known as poison pill...) so that the unreachable node may (if things aren't too badly broken) take its resources down, and stop the disk heartbeat, which would then allow the rest of the cluster to consider it having left the cluster safely, and migrate the resources. Not quite sure how much of a fundamental change this would be though... My suggestion for this would be to implement a full communications plugin module that sends packets through disk areas. If you do this right, then the communications will remain fully up for all purposes. We've had people start this effort in the past, but it's never been finished and all the bugs driven out AFAIK. Agreed. Since I can't make much headway with my other approach(es)... (and since having thought about it, they're certainly very much inferior to a full disk-based comms)... I happen to have a little time on my hands this month, and an itch to do some hacking. Can anybody point me to the patch(es) with whatever code we have around this? Is anybody else coding on this right now? Thanks Yan ___ Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
[Linux-ha-dev] Re: [Linux-HA] Recovering from unexpected bad things - is STONITH the answer?
Kevin Tomlinson wrote: On Tue, 2007-11-06 at 10:25 -0700, Alan Robertson wrote: We now have the ComponentFail test in CTS. Thanks Lars for getting it going! And, in the process, it's showing up some kinds of problems that we hadn't been looking for before. A couple examples of such problems can be found here: http://old.linux-foundation.org/developer_bugzilla/show_bug.cgi?id=1762 http://old.linux-foundation.org/developer_bugzilla/show_bug.cgi?id=1732 The question that comes up is this: For problems that should never happen like death of one of our core/key processes, is an immediate reboot of the machine the right recovery technique? The advantages of such a choice include: It is fast It will invoke recovery paths that we exercise a lot in testing It is MUCH simpler than trying to recover from all these cases, therefore almost certainly more reliable The disadvantages of such a choice include: It is crude, and very annoying It probably shouldn't be invoked for single-node clusters (?) It could be criticized as being lazy It shouldn't be invoked if there is another simple and correct method Continual rebooting becomes a possibility... We do not have a policy of doing this throughout the project, what we have is a few places where we do it. I propose that we should consider making a uniform policy decision for the project - and specifically decide to use ungraceful reboots as our recovery method for key processes dying (for example: CCM, heartbeat, CIB, CRM). It should work for those cases where people don't configure in watchdogs or explicitly define any STONITH devices, and also independently of quorum policies - because AFAIK it seems like the right choice, there's no technical reason not to do so. My inclination is to think that this is a good approach to take for problems that in our best-guess judgment shouldn't happen. I'm bringing this to both lists, so that we can hear comments both from developers and users. Comments please... I would say the right thing would depend on your cluster implementation and what is consider the right thing to do for the applications that the cluster is monitoring. I would propose that this action should be administrator configurable. From a user point of view with the cluster that we are implementing we would expect any cluster failure (internal) to either get itself back and running or just send out an alert Help me. im not working... as we would want our applications to continue running on the nodes. ** We dont want a service outage just because the cluster is no longer monitoring our applications. ** We would expect to get a 24x7 call out. Sev1 and then logon to the cluster and see what was happening. (configured alerting) Our applications only want a service outage if the node itself has issues not the Cluster.. Here's the issue: The solution as I see it is to do one of: a) reboot the node and clear the problem with certainty b) continue on and risk damaging your disks. c) write some new code to recover from specific cases more gracefully and then test it thoroughly. d) Try and figure out how to propagate the failure to the top layer of the cluster, and hope you get the notice there soon enough so that it can freeze the cluster before the code reacts to the apparent failure and begins to try and recover from it. In the current code, sometimes you'll get behavior (a) and sometimes you'll get behavior (b) and sometimes you'll get behavior (c). In the particular case described by bug 1762, failure to reboot the node did indeed start the same resource twice. In a cluster where you have shared disk (like yours for example), that would probably trash the filesystem. Not a good plan unless you're tired of your current job ;-). I'd like to take most/all of the cases where you might get behavior (b) and cause them to use behavior (a). If writing correct code and testing it were free, then (c) would obviously be the right choice. Quite honestly, I don't know how to do (d) in a reliable way at all. It's much more difficult than it sounds. Among other reasons, it relies on the components you're telling to freeze things to work correctly. Since resource freezes happen at the top level of the system, and the top layers need all the layers under them to work correctly, getting this right seems to be the kind of approach you could make into your life's work - and still never get it right. Case (c) has to be handled on a case by case basis, where you write and test the code for a particular failure case. IMHO the only feasible _general_ answer is (a). There are an infinite number of things that can go wrong. So, having a reliable and general strategy to deal with the WTF's of the world is a good thing. Of course, for those cases where we have a (c) behavior would not