v1.5.2

Update Go version and logrus version to cover a potential denial of service vulnerability in logrus.Writer()

V1.5.1

Update go version as for CVE-2023-245

v1.5.0

When the proxysql cluster feature was activated scheduler update the epoch of the node owning the lock at each cycle.
This injects not needed writes that could create conflicts especially in an environment like k8.

solution: Refresh epoch after few cycles but inside the lockclustertimeout value ideally every 3/4 of the lockclustertimeout value.
Example:
scheduler loop every 2 seconds
lockclustertimeout = 600 seconds
lockrefreshtime = 450

Implementation:

add parameter lockrefreshtime
add check for sanity for the parameter lockrefreshtime to never be higher than 3/4 of lockclustertimeout
add a check in findLock method to return true only if lockrefreshtime has been exceeded
add proper information in log (INFO)

V1.4.3

Modify the way we ping the servers adding a timeout to avoid the tool to hang because node was not reachable.
We have add the option to set a ping timeout when testing the connections.

The problem solved here is that when pinging without timeout if the node is not reachable the action remain hang for very long time.
Adding the timeout will allow the process to exit and report the error.
Additional parameter to drive this has been added for PXC cluster/nodes and ProxySQL.
The two values are independent.
The timeout value cannot be larger than the checkTimeOut.
The default value for pingTimeout is 1 second expressed in milliseconds pingTimeout=1000

V1.4.2

Version to fix a small bug on lock file time calculation which was preventing the file to be remove correctly.

V1.3.5

In this release we have fixed the bug https://jira.percona.com/browse/PSQLADM-381
Additional:

• Improved verification of the SSL certificates. When path or the file do not exist a WARNING message is raised
• If SSL is in use and file/path is not found ERROR message is raised, allowing the scheduler to complete the cleanup process. Previously was FATAL and execution was interrupted.
• New wrapper function for exit message

V1.3.4

Bug fix release.
Small bug on autofailover settings. Auto failover was always enable despite setting it to 0 (false)
The application will now respect the setting and will raise a warning in the log, given excluding the auto-failover goes against the main PXC concept of providing high level of availability.

V1.3.3

Fixing a problem as describe in #50
The issue was that under high contention the object keeping track the running threads was having issue in decrementing correctly the number of active threads.
The adoption of a mutex on the methods:

• increment
• decrement
• report

Fixed the issue.
The effects of the issue was that also if all nodes were correctly reported back, the scheduler was exiting the checking loop only on timeout. That was not compromising the checks but was causing a delay in action.
Other fix are spelling corrections as for PR-51 from @venkatesh-prasad-v

v1.3.2

Fixing:

• issue 47
• minor bug on performance reporting
• readme

v1.3.1

Fixed:

• Fixed a logic bug in the sanity check that was printing a warning without reason.
• Fixed a IPv6 issue in parsing the IP when moving up and down a node
• Updated documentation about variable case sensitivity
• Modified default config to have INFO log level and not debug
• Updated documentation to clarify different log levels
• Modified error handling when loading config file to be print with [ERROR] tag