ec.encode original volume maybe delete fail

[Numblgw]

Describe the bug
ec.encode command maybe fail when execute delete original volume step
command:

weed shell -master=xxxx
lock
ec.encode -collection=xxxx -volumeId=xxxx
unlock

log output:

markVolumeReadonly 154 on 10.117.32.36:9334 ...
markVolumeReadonly 154 on 10.117.32.34:9334 ...
markVolumeReadonly 154 on 10.117.32.35:9334 ...
........
........
delete volume 154 from 10.117.32.36:9334
delete volume 154 from 10.117.32.35:9334
delete volume 154 from 10.117.32.35:9334

10.117.32.35 delete execute repeated, 10.117.32.34 not delete

System Setup
any cluster with more than 2 replica, for example: -defaultReplication=020

Screenshots

Additional context
i think this caused by the shared memory of the locations slice.While iterating through the locations and deleting the original volume, the master client will also update the members in the locations.

reproduce: add a sleep within the loop. file: weed/shell/command_ec_encode.go, line: 185

[chrislusf]

thanks for the detailed report and insights!