Troubleshooting

Something is wrong with the cluster. This chapter is sorted by what you can see rather than by what is broken, because you are looking for the cause.

Open Start from the symptom first. It lists what an administrator actually sees and, for each entry, names the page that covers it. Start from the symptom also names the readings that look like faults but come from the cluster working as designed. Knowing these readings is the shortest way to avoid repairing something that repairs itself.

The pages that follow are grouped by the part of the cluster the trouble is in. Cluster health problems covers chunks the cluster reports as missing, endangered, or below their Redundancy Level. Master Server problems covers a Master Server that will not start, metadata that will not load, and a group left without a Leader. Chunkserver and disk recovery covers failed disks, a Chunkserver that will not register, and rescuing single files. Client and mount problems covers mounts that will not start, operations that hang, and changes that appear late on another client. Windows Client problems PRO (PRO Edition only) covers the symptoms specific to that client: the drive that cannot be opened or is missing from Explorer. Network problems covers the failures that appear in the cluster as something else entirely.

Diagnostic tools answers which tool to use for which question. Contacting support covers what to gather before asking for help.

Monitoring covers reading the cluster's own numbers. This chapter covers repairing what those numbers show.