Oracle Testcontainers shm error on Kubernetes >= 1.30

[jonmcewen]

Scenario

• GitLab Runner deployed to Amazon EKS with Sysbox runtime.
• CI job running a Spring Boot test using testcontainers to start an Oracle XE database.

Issue

The above works with Sysbox v0.6.4 on Kubernetes 1.29, but fails with Sysbox v0.6.6 on Kubernetes 1.30, with the following error message from the Oracle database:

ORA-00600: internal error code, arguments: [ksmcsg: failed to get total shm available], [], [], [], [], [], [], [], [], [], [], []
ORA-27300: OS system dependent operation:Cannot open /proc/sys/kernel/shm failed with status: 75
ORA-27301: OS failure message: Value too large for defined data type
ORA-27302: failure occurred at: sskgmgtssa_1

Diagnosis

My investigation so far suggests that the problem was introduced in cri-o 1.30.

I have tried the following in an attempt to diagnose:

• Sysbox v0.6.6 on Kubernetes 1.30 - fails as described above
• Sysbox v0.6.4 on Kubernetes 1.30 - fails with the same error
• Sysbox v0.6.4 on Kubernetes 1.29 - works (test passes)
• Kubernetes 1.30 (containerd), with no sysbox - works

Minimal project

I have created a minimal Maven/Spring Boot project that can be used to recreate the issue.

[jonmcewen]

The only likely cause I can see is this, but I can't understand why: nestybox/cri-o@f7c3a7f

There is a condition added round the chown of shmPath (line 578)

[jonmcewen]

@ctalledo please could you have a look at this when you have time (or suggest someone else who could)?. Thanks

[ctalledo]

Fix PR was merged on 03/27; will be present in the next Sysbox release (i.e., after v0.6.6)