Details
Description
Currently mariabackup can fail reading InnoDB redo log during backup due to two reasons:
1) Redo log block was incompletely written by server or damaged. In this case mariabackup tries to re-read log block several times before fail.
2) Redo log block was overwritten, as it is currently implemented as circular buffer, and mariabackup can read it slower then it is written.
In both cases there will be the same error message in backup log. In the second case the workaround is to increase InnoDB redo log size. But users do not know about such workaround as they don't have clear enough error message.
mariabackup uses InnoDB function to read redo log: log_t::file::read_log_seg(). Each log block contains it's log block number which can be converted in log block LSN. The function returns error if log block LSN does no match requested LSN. In mariabackup we could get log block number from log_sys.buf and check if "read log block number LSN" - "requested LSN" > "log size", then this is the second case, and we could output error message with advise to increase redo log size.
Attachments
Issue Links
- relates to
-
MDEV-18611 mariabackup silently ended during xtrabackup_copy_logfile()
- Closed
-
MDEV-19492 Mariabackup hangs if table populated with INSERT... SELECT while it runs
- Stalled
- mentioned in
-
Page Loading...