hcq/oceanbase
hcq/oceanbase
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

detect standby read when read uncleanout data

Browse Source
This commit is contained in:
Handora 2025-01-03 08:45:04 +00:00 committed by ob-robot
parent d5601c32f8
commit dd81a4ce6a
1 changed files with 16 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
src/storage/memtable/mvcc/ob_mvcc_iterator.cpp
View File

@ -276,14 +276,28 @@ int ObMvccValueIterator::lock_for_read_inner_(const ObQueryFlag &flag,
if (OB_FAIL(try_cleanout_tx_node_(iter))) {
TRANS_LOG(WARN, "cleanout tx state failed", K(ret), KPC(value_), KPC(iter));
}
// NB: We rely on the row_scn and state on the tx node if we really can
// read from the tx node. So if the tx node is not cleanout, we must
// Tip1: We rely on the row_scn and state on the tx node if we really
// can read from the tx node. So if the tx node is not cleanout, we must
// wait until the tx node is written back its state.
if (0 == (++counter) % 10000
&& REACH_TIME_INTERVAL(1_s)) {
TRANS_LOG(WARN, "waiting for the iter to be cleanout", K(ret),
KPC(iter), K(lock_for_read_arg), KPC(value_), KPC(ctx_));
}
// Tip2: In the transfer scenario, the tx_table and data at the src and
// dest are independent. Therefore, it is possible to use the src's data
// with the dest's tx_table. In the case, when reading from the standby,
// the tx_data may have already been updated while the data cannot be
// cleanout. Thus, it is necessary to detect and avoid such situations.
if (1 == counter % 10000
&& !MTL_TENANT_ROLE_CACHE_IS_PRIMARY_OR_INVALID()) {
ctx_->is_standby_read_ = true;
TRANS_LOG(WARN, "encounter standyby read with uncleanout data", K(ret),
KPC(iter), K(lock_for_read_arg), KPC(value_), KPC(ctx_));
}
usleep(10); // 10us
}
version_iter_ = iter;