Storage Manager Troubleshooting

This page describes how to troubleshoot Storage Manager issues and common errors.

Out of memory (OOM) issues

The following list provides some causes and possible resolutions to OOM events:

OOM from EOI process at writedown

An interval burst can cause the EOI process to OOM at writedown. In this case, sorting and attribute applications need extra RAM.

Recommendations

• Enable KXI_SM_EOI_SORT and KXI_SM_EOI_SORT_LIMIT (not enabled by default).
• Increase the memory available to Storage Manager.
• Lower pctMemThreshold to trigger the emergency EOI sooner.

Tables without emergency EOI protection

The basic / splayed / splayed_mem table types do not have emergency EOI protection, which can potentially lead to OOM errors.

Recommendations

• Use these table types only for static, non-increasing datasets.
• Use partitioned tables for tables that can outgrow memory.

Applying parted attribute

The size of the column on disk affects any attempt to apply a parted (p#) attribute.

Storage Manager memory

The Storage Manager requires at least twice the memory allocated to column disk size. For more information on cluster sizing, see Sizing Guide.

Recommendations

• Ensure the memory allocated to the Storage Manager is sufficient for the largest projected table column.
• Shard the dataset.

Pods sharing a node without correct requests or limits

For example, a Data Access Process (DAP) and the Storage Manager might share a node. If a large query runs on the DAP during a Storage Manager sort, this can cause OOM errors.

Recommendations

• Set Kubernetes (k8s) requests and limits so pods are isolated and correctly sized.

Next steps

For more information on cluster sizing, see Sizing Guide.