comdb2
comdb2 copied to clipboard
bloomberg
Bloomberg's distributed RDBMS
null=yes should not be needed in this case (see first commit) Adds csc2 and ddl support Provided that: * The col the sequence is added on is the leftmost field...
After RENAME, the file-version-cache would hold onto the old file versions, which would cause stat-size to misreport the table size. This patch fixes it.
Snapshot-lite is like snapshot but allows for retries on commit failures. Each retry is run in a new snapshot. --- [1](https://github.com/bloomberg/comdb2/blob/94ad0f86407e2989deb62b81994c4538ed8ff75f/db/sqlglue.c#L12932), [2](https://github.com/bloomberg/comdb2/blob/94ad0f86407e2989deb62b81994c4538ed8ff75f/db/sqlglue.c#L13039), [3](https://github.com/bloomberg/comdb2/blob/94ad0f86407e2989deb62b81994c4538ed8ff75f/db/fdb_bend_sql.c#L290), [4]( https://github.com/bloomberg/comdb2/blob/94ad0f86407e2989deb62b81994c4538ed8ff75f/db/watchdog.c#L97)
Bug Description: During parallel execution, the sqlengines executing in parallel don't respect maxquerytime set by client Steps to reproduce SET maxquerytime 3 select sleep(6) union all select sleep(6) Client returns...
RDSICDBDBA-1342 describes the issue while migrating from a 16 node cluster to a 24 node cluster. The cause is a overflow resulting in memory corruption. This patch replaces the static...
Introduce sc_add_drop_race test which executes multiple concurrent schema- changes against random, sometimes overlapping tables and fix the issues exposed by this test: * sc_tables_hash prevents concurrent schema-changes on different tables...
The fix is straightforward: check to see if the transaction has already written a blkseq record before returning a verify error.
The fix is straightforward: check to see if the transaction has already written a blkseq record before returning a verify error. This is the 8.0 version of https://github.com/bloomberg/comdb2/pull/4505