How to recover failed segments in batches

Products

• VMware Tanzu Greenplum

Issue/Introduction

Parallel recovery on multiple segments may consume more system resources like IO bandwidth, which could possibly affect system performance if there are many other jobs running. We are able to control the number of segments for recovery at a time.

Environment

Product Version: 5.0

Resolution

Refer to the video below for a demonstration:



Firstly, we can use gprecoverseg -o /tmp/recovery.lst to generate a list file /tmp/recovery.lst, which contains the information of all the failed segments, one line for each failed segment. We can edit this file, for example, delete some lines to control the number of segments to be recovered at present.

Now we can use gprecoverseg -i /tmp/recovery.lst to recover the failed segments that we specified.

Then we can do such operations repeatedly for the remaining failed segments.