Hi Vu, Ack from me.
Thanks, Zoran -----Original Message----- From: Vu Minh Nguyen [mailto:vu.m.ngu...@dektech.com.au] Sent: den 27 december 2017 11:32 To: Zoran Milinkovic <zoran.milinko...@ericsson.com> Cc: opensaf-devel@lists.sourceforge.net; Vu Minh Nguyen <vu.m.ngu...@dektech.com.au> Subject: [PATCH 1/1] imm: fix immnd restarted during SYNC finalize [#2749] The call to saImmOiImplementerSet and saImmOiImplementerClear are allowed during sync. Therefore, there is possibility the messages of these calls arrived at the sync-client after finalizeSync was sent from IMMND coord but before finalizeSync arrived at sync-client (over fevs). In that case, the implementer record stored in sImplementerVector database will then be different with the implementer record in finalizeSync and this mismatch caused the IMMND sync-client restart with current implementation. This ticket adds a check to ignore that implementer record in finalizeSync as it is obsolete instead of terminating the sync-client. --- src/imm/immnd/ImmModel.cc | 32 ++++++++++++++++++++++++++++++-- 1 file changed, 30 insertions(+), 2 deletions(-) diff --git a/src/imm/immnd/ImmModel.cc b/src/imm/immnd/ImmModel.cc index 16fd4d8..fcd354e 100644 --- a/src/imm/immnd/ImmModel.cc +++ b/src/imm/immnd/ImmModel.cc @@ -19476,10 +19476,38 @@ SaAisErrorT ImmModel::finalizeSync(ImmsvOmFinalizeSync* req, bool isCoord, "finalizeSync, ignoring", info->mImplementerName.c_str(), info->mId); continue; + } else if (ii->id == 0 && ii->nodeId == 0 && ii->mds_dest == 0) { + /* Implementer set is requested just before finalizeSync, and it + * arrives on nodes after finalizeSync sends implementer info with + * all 0. So, ignore the implementer record in finalizeSync as it + * is obsolete. + */ + LOG_WA("implementerSet '%s' id: %u has arrived " + "after fynalizeSync broadcast and " + "before finalizeSync arrived to the sync-client, ignoring", + info->mImplementerName.c_str(), info->mId); + continue; + } else if (info->mId == 0 && info->mNodeId == 0 && + info->mMds_dest == 0) { + /* Implementer clear is requested just before finalizeSync, and it + * arrives on nodes after finalizeSync sends implementer info with + * old set. So, ignore the implementer record in finalizeSync as it + * is obsolete. + */ + LOG_WA("implementerClear '%s' id: %u has arrived " + "after fynalizeSync broadcast and " + "before finalizeSync arrived to the sync-client, ignoring", + info->mImplementerName.c_str(), ii->id); + continue; } else { LOG_WA( - "Invalid implementer detected at sync client by finalizeSync -" - " exiting"); + "Mismatched implementer '%s' detected at sync client by" + "finalizeSync, info: id(%u, %u), nodeid(0x%x, 0x%x)," + "mdsdest(%" PRIx64", %" PRIx64 ") - exiting", + info->mImplementerName.c_str(), + info->mId, ii->id, + info->mNodeId, ii->nodeId, + (uint64_t)info->mMds_dest, (uint64_t)ii->mds_dest); err = SA_AIS_ERR_BAD_OPERATION; goto done; } -- 1.9.1 ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Opensaf-devel mailing list Opensaf-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/opensaf-devel
