hcq/oceanbase
hcq/oceanbase
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[BUG] torelant cannot find participant error

Browse Source
This commit is contained in:
Handora 2024-05-23 06:25:57 +00:00 committed by ob-robot
parent f5f654c73a
commit f02af51238
1 changed files with 22 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
src/storage/tx/ob_tx_2pc_msg_handler.cpp
View File

@ -792,7 +792,17 @@ int ObPartTransCtx::handle_tx_2pc_prepare_redo_resp(const Ob2pcPrepareRedoRespMs
if (OB_FAIL(set_2pc_request_id_(msg.request_id_))) {
TRANS_LOG(WARN, "set request id failed", KR(ret), K(msg), K(*this));
} else if (OB_FAIL(find_participant_id_(msg.sender_, participant_id))) {
TRANS_LOG(ERROR, "find participant failed", KR(ret), K(msg), K(*this));
if (0 == exec_info_.participants_.count()) {
// It may be possible that when the coordinator switches to the new
// leader, compensates the abort log while it may have already broadcasted
// the prepare redo requests by the old leader. And during the paxos of
// the abort log, it may receive the prepare redo response and has no
// participants list to handle the response, so we need tolerate it here.
ret = OB_SUCCESS;
TRANS_LOG(INFO, "find participant failed", KR(ret), K(msg), K(*this));
} else {
TRANS_LOG(ERROR, "find participant failed", KR(ret), K(msg), K(*this));
}
} else if (OB_FAIL(handle_2pc_resp(msg_type, participant_id))) {
TRANS_LOG(WARN, "handle 2pc response failed", KR(ret), K(msg), K(participant_id), K(*this));
}
@ -968,7 +978,17 @@ int ObPartTransCtx::handle_tx_2pc_abort_resp(const Ob2pcAbortRespMsg &msg)
if (OB_FAIL(set_2pc_request_id_(msg.request_id_))) {
TRANS_LOG(WARN, "set request id failed", KR(ret), K(msg), K(*this));
} else if (OB_FAIL(find_participant_id_(msg.sender_, participant_id))) {
TRANS_LOG(ERROR, "find participant failed", KR(ret), K(msg), K(*this));
if (0 == exec_info_.participants_.count()) {
// It may be possible that when the coordinator switches to the new
// leader, compensates the abort log while it may have already broadcasted
// the abort requests by the old leader. And during the paxos of the
// abort log, it may receive the abort response and has no participants
// list to handle the response, so we need tolerate it here.
ret = OB_SUCCESS;
TRANS_LOG(INFO, "find participant failed", KR(ret), K(msg), K(*this));
} else {
TRANS_LOG(ERROR, "find participant failed", KR(ret), K(msg), K(*this));
}
} else if (OB_FAIL(handle_2pc_resp(msg_type, participant_id))) {
TRANS_LOG(WARN, "handle 2pc response failed", KR(ret), K(msg), K(participant_id), K(*this));
}