Document Type | Troubleshooting
Category | Backup/Recovery
Applicable Product Versions | 6FS07, 7FS02
Document Number | TBATS010
Issue
When starting Standby with the tbboot failover command, a delay occurs during the Crash Recovery process. In the slog, delays are observed in the log section THREAD #0 adjust ckpt rba ~ REDO THREAD #0 CKPT updated to.
Cause
Some lscan logic is implemented to read blocks only in units of 8 blocks, which is the value of LSCAN_MULTI_READ_BLKCNT.
As a result, recovery is performed by reading the redo log in 8-block increments, causing the overall Crash Recovery to proceed at a very slow speed.
When this symptom occurs, you can confirm from the ilog logs below that block reads are limited to blkcnt=8.
07/31 10:47:22.6 0-0937 rv Iscan:354 lscan read block, one of logfiles in group #32, blkno-3455073, blkcnt=8 07/31 10:47:22.5 0-0937 rv Iscan:354 lscan read block, one of logfiles in group #32, blkno-3459001, blkcnt=8 07/31 10:47:22.6 0-8937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3458009, blkcnt=8 07/31 10:47:22.6 0-0937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459097, blkcnt=8 07/31 10:47:22.6 0-0937 rv 1scan:354 lscan read block. one of logfiles in group #32, blkno-3459105, blkcnt=8 07/31 10:47:22.5 0-0937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459113, blkcnt=8 07/31 10:47:22.6 0-0937 rv 1scan:354 lscan read black. one of logfiles in group #32, blkno-3459121, blkcnt=8 07/31 10:47:22.6 0-0937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459129, blkcnt=8 07/31 18:47:22.6 0-9937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459137, blkcnt=8 07/31 10:47:22.6 0-0937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459145, blkcnt=8 07/31 10:47:22.6 0-9937 rv Iscan:354 lscan read block. one of logfiles in group #32, blkno-3459153, blkcnt=8 07/31 10:47:22.7 0-0937 rv Iscan:354 lscan read block. one of logfiles in group a32, blkno-3459161, blkcnt=8
Solutions
Applying patch 308619 modifies the process to read in units of 2048 blocks (default value), allowing Crash Recovery to be performed more efficiently in this case.
By applying this patch, a large volume of 2048 blocks is read at once, improving recovery performance without unnecessary repeated operations.
CautionThe patch should be applied through technical support provided by Tmax Tibero.
After applying the patch, you can verify that blocks are being read in large volumes with blkcnt=2048.
08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=1, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=2049, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=4097, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=6145, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=8193, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=10241, blkcnt=2048 08/03 09:40:52.2 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=12289, blkcnt=2048 08/03 09:40:52.3 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=14337, blkcnt=2048 08/03 09:40:52.3 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=16385, blkcnt=2048 08/03 09:40:52.3 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=18433, blkcnt=2048 08/03 09:40:52.3 0-0138 rv_lscan:353 lscan read block. one of logfiles in group #4, blkno=20481, blkcnt=2048
Note