Conversation
markhannum
force-pushed
the
blkseq_commitlsn
branch
from
d04028e to
e1e7f24
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
truncatesc_offline_generated **quarantined**
sc_resume_logicalsc_generated **quarantined**
reco-ddlk-sql **quarantined**
sc_parallel_logicalsc_generated
consumer_non_atomic_default_consumer_generated **quarantined**
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
consumer_non_atomic_default_consumer_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
markhannum
force-pushed
the
blkseq_commitlsn
branch
from
e1e7f24 to
22c1b8d
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
scindex_logicalsc_generated [failed with core dumped]
consumer_non_atomic_default_consumer_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
markhannum
force-pushed
the
blkseq_commitlsn
branch
3 times, most recently
from
9ce4ffe to
d4ba33c
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Error ⚠.
Regression testing: Success ✓.
The first 10 failing tests are:
consumer_non_atomic_default_consumer_generated **quarantined**
unifiedcancel **quarantined**
truncatesc_offline_generated [timeout] **quarantined**
reco-ddlk-sql [timeout] **quarantined**
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Error ⚠.
Regression testing: Success ✓.
The first 10 failing tests are:
socksql_master_swings
cldeadlock
consumer_non_atomic_default_consumer_generated **quarantined**
sc_transactional_rowlocks_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
markhannum
force-pushed
the
blkseq_commitlsn
branch
3 times, most recently
from
b288f1c to
ff4d085
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
scindex_logicalsc_generated
consumer_non_atomic_default_consumer_generated **quarantined**
sc_transactional_rowlocks_generated **quarantined**
queuedb_rollover [timeout] **quarantined**
sc_truncate_lockorder_generated [timeout] **quarantined**
reco-ddlk-sql [timeout] **quarantined**
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
sc_truncate_multiddl_generated [db unavailable at finish] **quarantined**
socksql_master_swings
queuedb_rollover_noroll1_generated **quarantined**
consumer_non_atomic_default_consumer_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
| /* Prepare failed to reach a majority of the cluster: fail the txn */ | ||
| dist_txn_abort_write_blkseq(thedb->bdb_env, bskey, bskeylen); | ||
| trans_abort(iq, parent_trans); | ||
| dist_txn_abort_write_blkseq(thedb->bdb_env, bskey, bskeylen, NULL, 0); |
There was a problem hiding this comment.
should we call this before releasing the locks, and use parent_trans?
There was a problem hiding this comment.
trans_abort releases all locks and will emit the 'dist-abort' record. The code is a bit lazy in forcing all non-durable prepares to fall back to the 'latest-commit-is-durable' case.
I wrote it this way because only prepared transactions require an explicit 'abort' record, and because we don't currently have a way to convey the LSN of that abort record to toblock.
markhannum
force-pushed
the
blkseq_commitlsn
branch
2 times, most recently
from
8818b08 to
ba780e3
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
truncatesc_offline_generated [failed with core dumped] **quarantined**
sc_resume_logicalsc_generated **quarantined**
consumer_non_atomic_default_consumer_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
lostwrite
consumer_non_atomic_default_consumer_generated **quarantined**
phys_rep_tiered_firstfile_generated [timeout]
reco-ddlk-sql [timeout] **quarantined**
markhannum
force-pushed
the
blkseq_commitlsn
branch
from
ba780e3 to
9387969
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Error ⚠.
Regression testing: Success ✓.
The first 10 failing tests are:
consumer_non_atomic_default_consumer_generated **quarantined**
longreq_stats
markhannum
force-pushed
the
blkseq_commitlsn
branch
2 times, most recently
from
556b21b to
3363f10
Compare
Signed-off-by: Mark Hannum <mhannum@bloomberg.net>
Signed-off-by: Mark Hannum <mhannum@bloomberg.net>
markhannum
force-pushed
the
blkseq_commitlsn
branch
from
3363f10 to
6045bbd
Compare
roborivers
left a comment
roborivers
left a comment
There was a problem hiding this comment.
Cbuild submission: Success ✓.
Regression testing: Success ✓.
The first 10 failing tests are:
sc_resume_logicalsc_generated **quarantined**
timepart_retro
consumer_non_atomic_default_consumer_generated **quarantined**
reco-ddlk-sql [timeout] **quarantined**
3 participants
This PR changes the durability algorithm used for blkseq replays. If we detected a replay previously, the master would return 'this is durable' only if the most recent commit was durable. This instead records a transactions's blkseq->commitlsn mapping in the blkseq_commitlsns table. This allows us to determine durability using the transaction's actual commit-lsn, falling through to the previous logic if we cannot find it.