Releases: ClusterLabs/striker
Version 2.0.0 beta - Release Candidate 24
Small bump to RC23 that fixes a bug in changing passwords on an Anvil! with a new password that has spaces or exclamation marks removed. Also added better messaging to tell the user that this can happen (or has happened, in striker-change-password
).
Version 2.0.0 beta - Release Candidate 23
This is a minor update to RC22 that resolves issue #125 where nodes with interfaces that come up with interface names other than ethX
would not be removed properly, causing the udev rules file to be mangled on reboot during an Install Manifest run.
Version 2.0.0 beta - Release Candidate 22
Minor update over RC21.
- Fixed a bug where rsync wrappers were being left around, which is a security issue as well as caused problems when the wrapper for a target is written by a root user, later the password changes and apache can't rewrite the wrapper with the new password (and the dashboard hadn't rebooted).
- Fixed a bug with how passwords are altered to work with IPMI interfaces during Install Manifest runs.
- Updated
striker-installer
to use the system's UUID and removed the preset--host-uuid
fromstriker-installer.example
.
Version 2.0.0 beta - Release Candidate 21
This release fixes a couple significant issues discovered in RC20 testing. Upgrading is advised.
The core issue is that the historical data recorded for LSI-controller attached physical disks were storing their data in the historical database schema out of order, causing a rebuilt dashboard's public schema to be populated with incorrect data. This would be corrected in the next scan as the erroneous data would appear to be drives that vanished, but it left the historical data incorrect. This release fixes this problem.
The striker-update
tool will detect this fault and correct it. if you have a Striker release newer than TC2, you can update without a reinstall.
Other changes;
striker-update
now stops apache to disable the Striker UI while an update is running. This was done to minimize errors that might be displayed to the user if they tried to use the web UI while updates were under way.- Updated
scan-ipmitool
to ignore theBBU
sensor as it appears to frequently report deviations that do not actually exist. The BBU (or FBU) is monitored directly byscan-storcli
so there is no loss in coverage. - Fixed a bug where notification recipient alert level was not preserved in some cases when a striker dashboard was rebuild and the
recipients
table had to resynchronize. - Fixed a bug where a lock request variable would be duplicated if a dashboard was rebuilt and the newly built dashboard tried to resync before its peer.
- Several minor display/string bugs corrected.
Version 2.0.0 beta - Release Candidate 19
This is a minor update to RC19 that fixes a minor display error in anvil-generate-iso
and adds proper support for Dell-based Anvil! systems using PERC controllers.
Version 2.0.0 beta - Release Candidate 19
This is a minor update to RC18 that fixes bugs found in RC18 testing.
Notable fixes;
- Saving archiving of database data is now disabled by default.
- Fixed a bug where servers that returned an IPMI sensor limit as '
inf
' would cause a database query error. - Fixed a bug where
scan-storcli
would throw warnings when the hex sector count was >32bit. - Fixed how thermal load sheds are handled so that cool-down boot delays are handled better.
Version 2.0.0 beta - Release Candidate 18
This commit includes many small fixes from testing of RC17;
Bug fixes of note;
- Fixed a bug(?) where, if a node was rebuilt and then later used to upload a file to the media library, it would fail because of the old SSH fingerprint not matching the new one. Now, we remove the offending key so that it starts working again on subsequent uploads. The is a questionable approach for security reasons and may be changed in a future release. For now, the logic is that the BCN is what dashboard -> node comms happen on and it is supposed to be secured and isolated.
- NOTE: This change is worth input - If you feel this was the wrong choice, please comment.
- Changed the default drbd c-plan-ahead to '1' and the default max-buffers to 8k on the advice of LINBIT staff to make DRBD 8.4 adjust the resync rate according to load better.
- Fixed a problem where trying to choose a CPU count for a server with a zero-pre-padded value stored as-is in the XML definition, which caused virsh to default to a single core.
- Fixed a bug in scan-storcli where a drive group wouldn't map to a controller UUID when the drive group vanished causing an error and exit.
- Fixed a bug in Get->server_host() where it would error out trying to check what servers were running on the peer whem the peer was offline.
- Fixed a bug in ScanCore where the '-v...' switches were not being honoured properly.
- Updated ScanCore's thermal "cool down" timers to count both thermal shutdowns and load sheds in the previous six hours.
- Fixed a bug where, if the peer was offline and the local node was in thermal warning, it would try to load shed.
- Upped the verbosity of logging in ScanCore's 'avoid_duplicate_run' function after a (so far unreproducable) bug where two copies of ScanCore managed to run at the same time.
Version 2.0.0 beta - Release Candidate 17
This is a minor update;
Improvements;
- The major update in this release is that network link drops/recovers now influence the node's health.
- Added a ping check before connecting to a database to speed up processing when a ScanCore DB is lost.
- Added a check before running queries against databases to detect when a target DB has failed so that ScanCore doesn't hang on a stale DB filehandle.
Version 2.0.0 beta - Release Candidate 16
This release is mainly a bug-fix and minor enhancement update.
Key changes;
- The biggest change is improved handling of servers that fail to stop within two minutes and enter a 'failed' state' Messaging to the user has been significantly improved. Striker allows a failed server to be forced off and the state cleared. Finally,
scan-clustat
will clear the failed state of a server once it is powered off (ie: from within the OS or after updates finished installing on the guest, as often happens with windows). - New servers will no longer be destroyed on reboot, making it easier for users to interact with the server early in the boot process.
- Fixed a bug where trying to use the APC UPSes to power cycle or power off a rack while the Anvil! is offline was throwing an error.
- Improved
anvil-safe-start
's handling of storage that is in the 'disabled' state.
Version 2.0.0 beta - Release Candidate 15
Bug fixes;
- Fixed a bug in
striker-update
where updating a peer dashboard that is offline failed to actually update. - Removed support for Solaris based guests because, as of now, Solaris 11.3 fails to install with no clear solution in sight.
Enhancement
- This release adds the new "don't alert until X bad scans" logic to
scan-ipmitool
,scan-apc-ups
andscan-apc-pdu
. This is similar to the earlier change toscan-ipmitool
which used to no alarm until the second missed scan, but allows for a configurable number of misses before generating the alarm. - Added the alias '
--anvil
' to act like '--anvil-name
' instriker-update
.