gpinitstandby Fails after Expansion during Resynchronize (using -n)

Products

• VMware Tanzu Greenplum

Issue/Introduction

Symptoms:

After a cluster expansion - standby resynchronize (gpinitstandby -n) fails with error referring to unexpected location at standby master.

You can notice the error is pointing towards mirror directory at standby master as shown below:

[gpadmin@mdw]$ gpinitstandby -n
20140620:17:50:10:030748 gpinitstandby:mdw:gpadmin-[INFO]:-Validating environment and parameters for standby initialization...
20140620:17:50:10:030748 gpinitstandby:mdw:gpadmin-[ERROR]:-Parent directory /data1/mirror does not exist on host smdw
20140620:17:50:10:030748 gpinitstandby:mdw:gpadmin-[ERROR]:-This directory must be created before running gpactivatestandby
20140620:17:50:10:030748 gpinitstandby:mdw:gpadmin-[INFO]:-Trying to rollback changes that have been made...
20140620:17:50:10:030748 gpinitstandby:mdw:gpadmin-[ERROR]:-Error initializing standby master: Parent directory /data1/mirror does not exist

Environment

Cause

On GPDB versions before 4.2.8.1 gpinitstandby expects standby master to have highest dbid. However after the expansion, one of the newly added mirrors will/would acquire the highest dbid and therefore the location by the standby is pointing to that mirror location.

 

Resolution

• Recreate the standby and add it again back again.

Refer GPDB admin guide for more information or help on gpinitstandby execution.

gpinitstandby -r
gpinitstandby -s

• This issue has been fixed in 4.2.8.1 and later and should not show up in these versions
• If possible do upgrade to the fixed release before expanding the cluster.