Skip to content

1.5.7
Compare
Choose a tag to compare
@sungeun-kim sungeun-kim released this 09 Jul 09:05
· 653 commits to master since this release
WDRBD_1.5.7
0178bea

Key points

passthrough

The existing pass_on policy downgrades the disk state to inconsistent when disk I/O error occurs. Another I/O error policy, diskless mode, takes an action to overcome disk I/O errors through I/O redirection, but it also has to bear the burden of I/O redirection. We have added a passthrough policy logic to this by compromising the pass_on policy, because we do not want the operating node's disk state to be inconsistent, nor do we want to have the burden of I/O redirection.
WDRBD now does not detach a replication disk and does not downgrade the disk state when a disk I/O error occurs, simply mark the blocks that have I/O errors with OOS and take follow-up actions(automatic or manual resynchronization from SyncSource to SyncTarget).

AL Bottleneck improvement

DRBD runs AL and resync LRU together. These two structures seem to work well, but sometimes there is a bottleneck problem due to the dependency between the two structures. This leads to drastic performance degradation of local I/O performance. We tracked these difficult-to-reproduce problems for a long time, got some improvement points, and adjusted the logic. We will continue to observe and reinforce this issue in the future.

Apply EV Certificate

EV code signing for Windows 10/2016 has been added.

Release Note

Bug

[DW-1727] - [BSOD] submit_ee: Allocation of a bio failed
[DW-1732] - fix pausing progress during synchronization
[DW-1735] - [BSOD] When disconnecting during replication read_ee drbd_free_peer_reqs Memory is being released, PFN_LIST_CORRUPT bug check
[DW-1749] - [hang]fix OS hangs after disconnecting the replication network after IO occurs
[DW-1763] - max_bio_size datatype does not match
[DW-1773] - If EE_WRITE is set in peer_request flags and the WRITE request is terminated without a memory request
[DW-1775] - Sync does not finished at RECORD_RS_FAILED
[DW-1777] - "Opening /dev/null failed: No such device or address" error when running drbdadm create-md
[DW-1778] - [BSOD] _drbd_bm_lock () Occurs when b-> bm_task is referenced
[DW-1787] - The victim node disk state becomes outdate after the split brain is occurred and is not resolved
[DW-1791] - Memory leak for down/up
[DW-1794] - Log abnormal output when drbdlog command is executed
[DW-1801] - 1:N, Outdated when repeating down/up in secondary in multi-volume configuration
[DW-1804] - Failure of barrier ack pending timeout when changing from primary to secondary
[DW-1805] - Drbdsetup.exe terminates when receiving event of disk error during event monitor (drbdsetup events2)
[DW-1806] - SB is not resolved and some disks are retained after changing to Outdated or Inconsistent
[DW-1807] - Synchronization between A and C is not completed when A node is promoted to a C node while synchronizing with B with invalidate command.
[DW-1811] - During synchronization, after forced reboot, OOS is maintained
[DW-1812] - Paused in drbdsetup status Ahead / Behind after file copy
[DW-1816] - [BSOD] During the up/down operation in the low memory situation, the BSOD
[DW-1818] - [hang] Hang occurs when the replication volume of the target node is offline during synchronization
[DW-1826] - Multiple resource invalidate command does not complete synchronization of some resources
[DW-1829] - When calling drbd_destroy_connection () function, if inactive_ee peer_request exists, BSOD
[DW-1831] - [BSOD] fix null reference error on disk offline during replication
[DW-1832] - After primary disk failure, drbdsetup status Repeatedly No currently configured DRBD found Output

New Feature

[DW-500] - Applying EV Certificate - Windows 10/2016 driver
[DW-1721] - Setting up kernel dumps during installation
[DW-1755] - Improved disk io failure policy - passthrough

Task

[DW-1587] - Fix static analysis alerts
[DW-1704] - Resolve all build warnings

Assets 2
Loading