We are trying to upgrade our ubuntu cluster from Ubuntu 14.04 to Ubuntu 16.04.
We got a problem with percona xtradb cluster.
We use, on the old cluster, percona xtradb 5.6.41-28.28-1.trusty package from percona repo. If we upgrade ubuntu with this package, this package run fine on ubuntu 16.04. But if we upgrade to 5.6.41-28.28-1.xenial, all seem to work good (clustercheck is ok, sst is ok, etc…) but as soon as we direct some trafic to the upgraded node, it crash.
The only message we have in logs is:
Thread pointer: 0x2969f80
Attempting backtrace. You can use the following information to find out
where mysqld died. If you see no messages after this, something went
stack_bottom = 7f7f351ace38 thread_stack 0x40000
Trying to get some variables.
Some pointers may be invalid and cause the dump to abort.
Query (0): is an invalid pointer
Connection ID (thread ID): 75
When we try to execute manually some requests (select, inserts, etc…) we have no problem, but as soon as we add the real trafic it crash. We have try to log all request, and to replay requests concerned by the thread ID that crash, without any crash.
How is it possible that trusty version have no problem on xenial but xenial version have ? How can we upgrade ?