TudbuT/conduit
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

elaborate on RocksDB recovery modes in example-config.

Browse source 
Signed-off-by: Jason Volk <jason@zemos.net>
This commit is contained in:
Jason Volk 2024-03-24 18:08:42 -07:00 • committed by June
parent 24faf4dd26
commit f3bc87c4e6
1 changed files with 12 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

24
conduwuit-example.toml
View file

@ -361,21 +361,21 @@ url_preview_check_root_domain = false
# Defaults to false as this uses more CPU when compressing. # Defaults to false as this uses more CPU when compressing.
#rocksdb_bottommost_compression = false #rocksdb_bottommost_compression = false
# RocksDB WAL recovery mode # Database recovery mode (for RocksDB WAL corruption)
# #
# If the database becomes corrupted, an attempt at some form of recovery or at least opening it can be made # Use this option when the server reports corruption and refuses to start. Set mode 2 (PointInTime)
# using this config option. # to cleanly recover from this corruption. The server will continue from the last good state,
# several seconds or minutes prior to the crash. Clients may have to run "clear-cache & reload" to
# account for the rollback. Upon success, you may reset the mode back to default and restart again.
# Please note in some cases the corruption error may not be cleared for at least 30 minutes of
# operation in PointInTime mode.
# #
# In this event, provided that you have searched for any backups already, it's recommended to start by using PointInTime recovery mode. If this opens your database successfully, # As a very last ditch effort, if PointInTime does not fix or resolve anything, you can try mode
# you will want to immediately run the `clear-cache` database admin command and restart Conduwuit again setting this back to 1 (TolerateCorruptedTailRecords) # 3 (SkipAnyCorruptedRecord) but this will leave the server in a potentially inconsistent state.
# If no further issues arrise, your database should be okay now.
#
# As a very last ditch effort, if PointInTime does not fix or resolve anything, you can try SkipAnyCorruptedRecord but this
# has a good chance to do more damage than before
#
# TolerateCorruptedTailRecords is the default as generally tail records may be caused by unclean shutdowns, and any data here is likely
# federation data that can be re-retrieved by other servers again.
# #
# The default mode 1 (TolerateCorruptedTailRecords) will automatically drop the last entry in the
# database if corrupted during shutdown, but nothing more. It is extraordinarily unlikely this will
# desynchronize clients. To disable any form of silent rollback set mode 0 (AbsoluteConsistency).
# #
# The options are: # The options are:
# 0 = AbsoluteConsistency # 0 = AbsoluteConsistency