In one step upgrade, during the lock nodegroup. The timeout can happen
and it causes the upgrade failed. By retrying if the return code of
saImmOmAdminOperationInvoke_2() is SA_AIS_ERR_NO_OP, the lock is
considered as successfully.
---
src/smf/smfd/SmfAdminState.cc | 16 ++++++++++++++++
1 file changed, 16 insertions(+)
diff --git a/src/smf/smfd/SmfAdminState.cc b/src/smf/smfd/SmfAdminState.cc
index 90ae093c4..7b6155cd0 100755
--- a/src/smf/smfd/SmfAdminState.cc
+++ b/src/smf/smfd/SmfAdminState.cc
@@ -917,11 +917,13 @@ bool SmfAdminStateHandler::nodeGroupAdminOperation(
SaAisErrorT oi_rc = SA_AIS_OK;
SaAisErrorT imm_rc = SA_AIS_OK;
errno_ = SA_AIS_OK;
+ SaBoolT timeout_elapsed = SA_FALSE;
base::Timer adminOpTimer(smfd_cb->adminOpTimeout / kNanoMillis);
while (adminOpTimer.is_timeout() == false) {
TRACE("%s: saImmOmAdminOperationInvoke_2 time left = %" PRIu64,
__FUNCTION__, adminOpTimer.time_left());
+ timeout_elapsed = SA_FALSE;
imm_rc =
saImmOmAdminOperationInvoke_2(ownerHandle_, &nodeGroupName, 0,
adminOp, params, &oi_rc,
@@ -930,6 +932,20 @@ bool SmfAdminStateHandler::nodeGroupAdminOperation(
(imm_rc == SA_AIS_OK && oi_rc == SA_AIS_ERR_TRY_AGAIN)) {
base::Sleep(base::MillisToTimespec(2000));
continue;
+ } else if (imm_rc == SA_AIS_ERR_TIMEOUT) {
+ timeout_elapsed = SA_TRUE;
+ // Reset timeout
+ adminOpTimer.set_timeout_time(smfd_cb->adminOpTimeout / kNanoMillis);
+ continue;
+ } else if ((imm_rc == SA_AIS_ERR_NO_OP &&
+ timeout_elapsed == SA_TRUE)) {
+ // If an admin operation is already performed SA_AIS_ERR_NO_OP
+ // is returned. Treat this as OK, just log it and return
+ // operation success
+ LOG_NO("Admin op [%d] on [%s], return SA_AIS_ERR_NO_OP,"
+ "treated as OK", adminOp, nodeGroupName_s.c_str());
+ method_rc = true;
+ break;
} else if (imm_rc != SA_AIS_OK) {
LOG_NO(
"%s: saImmOmAdminOperationInvoke_2 Fail %s",
--
2.25.1
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel
