Closed fuero closed 2 months ago
So after a "little" more digging, I found the cause - the volume mount's block size.
This message was the dead giveaway, masquerading as an information level message:
There have been 256 misaligned log IOs which required falling back to synchronous IO. The current IO is on file /var/opt/mssql/data/master.mdf.
see Technet on this issue.
My Kubernetes cluster's CSI driver provisioned block devices with 16K block size, but SQL Server apparantly insists on block sizes <= 4K.
Reconfiguring the CSI driver for this particular PVC made the error vanish.
The container image crashes continuously with "Stack Overflow"
OS: Debian Bookworm Image: mcr.microsoft.com/mssql/server:2022-CU13-ubuntu-22.04@sha256:c4369c38385eba011c10906dc8892425831275bb035d5ce69656da8e29de50d8 Kernel: 6.1.0-22-amd64 Tested as well with 6.6.13+bpo, 6.7.12+bpo - same result, with the -cloud variants as well. k3s version v1.30.2+k3s1 (aa4794b3) Tested with an empty database Same deployment runs fine on Debian 11 with kernel