On Tue, Jun 7, 2011 at 9:45 AM, Wesley Bland <wbl...@eecs.utk.edu> wrote:
> To adress your concerns about putting the epoch in the process name > structure, putting it in there rather than in a separately maintained list > simplifies things later. > Not really concerned - I was just noting we had done it a tad differently, but nothing important. > > For example, during communication you need to attach the epoch to each of > your messages so they can be tracked later. If a process dies while the > message is in flight, or you need to cancel your communication, you need to > be able to find the matching message to the matching epoch. If the epoch > isn't in the process name, then you have to modify to the message header for > each type of message to include that information. Each process not only > needs to know what the current version of the epoch is from it's own > perspective, but also from the perspective of whomever is sending the > message. > But the epoch is process-unique - i.e., it is the number of times that this specific process has been started, which differs per proc since we don't restart all the procs every time one fails. So if I look at the epoch of the proc sending me a message, I really can't check it against my own value as the comparison is meaningless. All I really can do is check to see if it changed from the last time I heard from that proc, which would tell me that the proc has been restarted in the interim. > This is also true for things like reporting failures. To prevent duplicate > notifications you would need to include your epoch in all the notifications > so no one marks a process as failing twice. > I'm not sure of the relevance here. We handle this without problem right now (at least, within orcm - haven't looked inside orte yet to see what needs to be brought back, if anything) without an epoch - and the state machine will resolve the remaining race conditions, which really don't pertain to epoch anyway. > > Really the point is that by changing the process name, you prevent the need > to pack the epoch each time you have any sort of communication. All that > work is done along with packing the rest of the structure. > No argument - I don't mind having the value in the name. Makes no difference to me. > On Tuesday, June 7, 2011 at 11:21 AM, Ralph Castain wrote: > > Thanks for the explanation - as I said, I won't have time to really review > the patch this week, but appreciate the info. I don't really expect to see a > conflict as George had discussed this with me previously. > > I know I'll have merge conflicts with my state machine branch, which would > be ready for commit in the same time frame, but I'll hold off on that one > and deal with the merge issues on my side. > > > > On Tue, Jun 7, 2011 at 8:46 AM, Wesley Bland <wbl...@eecs.utk.edu> wrote: > > This could certainly work alongside another ORCM or any other fault > detection/prediction/recovery mechanism. Most of the code is just dedicated > to keeping the epoch up to date and tracking the status of the processes. > The underlying idea was to provide a way for the application to decide what > its fault policy would be rather than trying to dictate one in the runtime. > If any other layer wanted to register a callback function with this code, it > could do anything it wanted to on top of it. > > Wesley > > On Tuesday, June 7, 2011 at 7:41 AM, Ralph Castain wrote: > > I'm on travel this week, but will look this over when I return. From the > description, it sounds nearly identical to what we did in ORCM, so I expect > there won't be many issues. You do get some race conditions that the new > state machine code should help resolve. > > Only difference I can quickly see is that we chose not to modify the > process name structure, keeping the "epoch" (we called it "incarnation") as > a separate value. Since we aren't terribly concerned about backward > compatibility, I don't consider this a significant issue - but something the > community should recognize. > > My main concern will be to ensure that the new code contains enough > flexibility to allow integration with other layers such as ORCM without > creating potential conflict over "double protection" - i.e., if the layer > above ORTE wants to provide a certain level of fault protection, then ORTE > needs to get out of the way. > > > On Mon, Jun 6, 2011 at 1:00 PM, George Bosilca <bosi...@eecs.utk.edu>wrote: > > WHAT: Allow the runtime to handle fail-stop failures for both runtime > (daemons) or application level processes. This patch extends the > orte_process_name_t structure with a field to store the process epoch (the > number of times it died so far), and add an application failure notification > callback function to be registered in the runtime. > > WHY: Necessary to correctly implement the error handling in the MPI 2.2 > standard. In addition, such a resilient runtime is a cornerstone for any > level of fault tolerance support we want to provide in the future (such as > the MPI-3 Run-Through Stabilization or FT-MPI). > > WHEN: > > WHERE: Patch attached to this email, based on trunk r24747. > TIMEOUT: 2 weeks from now, on Monday 20 June. > > ------ > > MORE DETAILS: > > Currently the infrastructure required to enable any kind of fault tolerance > development in Open MPI (with the exception of the checkpoint/restart) is > missing. However, before developing any fault tolerant support at the > application (MPI) level, we need to have a resilient runtime. The changes in > this patch address this lack of support and would allow anyone to implement > a fault tolerance protocol at the MPI layer without having to worry about > the ORTE stabilization. > > This patch will allow the runtime to drop any dead daemons, and re-route > all communications around the holes in order to __ALWAYS__ deliver a message > as long as the destination process is alive. The application is informed > (via a callback) about the loss of the processes with the same jobid. In > this patch we do not address the MPI_ERROR_RETURN type of failures, we > focused on the MPI_ERROR_ABORT ones. Moreover, we empowered the application > level with the decision, instead of taking it down in the runtime. > > NEW STUFF: > > Epoch - A counter that tracks the number of times a process has been > detected to have terminated, either from a failure or an expected > termination. After the termination is detected, the HNP coordinates all > other process’s knowledge of the new epoch. Each ORTED will know the epoch > of the other processes in the job, but it will not actually store anything > until the epochs change. > > Run-Through Stabilization - When an ORTED (or HNP) detects that another > process has terminated, it repairs the routing layer and informs the HNP. > The HNP tells all other processes about the failure so they can also repair > their routing layers an update their internal bookkeeping. The processes do > not abort after the termination is detected. > > Callback Function - When the HNP tells all the ORTEDs about the failures, > they tell the ORTE layers within the applications. The application level > ORTE layers have a callback function that they use to inform the OMPI layer > about the error. Currently the OMPI errhandler code fills in this callback > function so it is informed when there is an error and it aborts (to maintain > the current default behavior of MPI). This callback function can also be > used in an ORTE only application to perform application based fault > tolerance (ABFT) and allow the application to continue. > > NECESSARY FOR IMPLEMENTATION: > > Epoch - The orte_process_name_t struct now has a field for epoch. This > means that whenever sending a message, the most current version of the epoch > needs to be in this field. This is a simple look up using the function in > orte/util/nidmap.c: orte_util_lookup_epoch(). In the orte/orted/orted_comm.c > code, there is a check to make sure that it isn’t trying to send messages to > a process that has already terminated (don’t send to a process with an epoch > less than the current epoch). Make sure that if you are sending a message, > you have the most up to date data here. > > Routing - So far, only the binomial routing layer has been updated to use > the new resilience features. To modify other routing layers to be able to > continue running after a process failure, they need to be able to detect > which processes are not currently running and route around them. The errmgr > gives the routing layer two chances to do this. First it calls delete_route > for each process that fails, then it calls update_routing_tree after it has > appropriately marked each process. Before either of these things happen the > epoch and process state have already been updates so the routing layer can > use this data to determine which processes are alive and which are dead. A > convenience function has been added to orte/util/nidmap.h called > orte_util_proc_is_running() which allows the ORTEDs to determine the status > of a process. Keep in mind that a process is not running if it hasn’t > started up yet so it is wise to check the epoch (to make sure that it isn’t > ORTE_EPOCH_MIN) as well to make sure that you’re actually detecting an error > and not just noticing that an ORTED hasn’t finished starting. > > Callback - If you want to implement some sort of fault tolerance on top of > this code, use the callback function in the errmgr framework. There is a new > function in the errmgr code called set_fault_callback that takes a function > pointer. The ompi_init code sets this to a default value just after it calls > orte_init (to make sure that there is an errmgr to call into). If you later > set this to a new function, you will get the callback to notify you of > process failures. Remember that you’ll need to handle any sort of MPI level > fault tolerance at this point because you’ve taken away the callback for the > OMPI layer. > > > > > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel >
