fixes bug with starting multiple group of scan servers #4445
Conversation
When trying to start and stop multiple groups of scan servers using the accumulo-cluster script the groups would interfere with each other. This change fixes that by making the the scan servers use the group in the instance id like compactors do. Ran into this while workiing on apache#4444 and pulled it out as a stand alone fix.
| @@ -141,7 +141,7 @@ function control_service() { | |||
| ACCUMULO_SERVICE_INSTANCE="" | |||
| [[ $service == "tserver" && ${NUM_TSERVERS:-1} -gt 1 ]] && ACCUMULO_SERVICE_INSTANCE=${inst_id} | |||
| [[ $service == "compactor" ]] && ACCUMULO_SERVICE_INSTANCE="${inst_id}_${5}" | |||
| [[ $service == "sserver" && ${NUM_SSERVERS:-1} -gt 1 ]] && ACCUMULO_SERVICE_INSTANCE=${inst_id} | |||
| [[ $service == "sserver" ]] && ACCUMULO_SERVICE_INSTANCE="${inst_id}_${5}" | |||
There was a problem hiding this comment.
We could simplify the code a bit more and only use the sserver1_<group> or compactor1_<queue> patterns when there's more than one of each type.
ACCUMULO_SERVICE_INSTANCE=""
# Only increment the service name when more than one service is desired.
[[ last_instance_id -gt 1 ]] && ACCUMULO_SERVICE_INSTANCE="${inst_id}"
[[ $service == "compactor" || $service == "sserver" ]] && ACCUMULO_SERVICE_INSTANCE="${ACCUMULO_SERVICE_INSTANCE}_${5}"
There was a problem hiding this comment.
We could simplify the code a bit more and only use the sserver1_ or compactor1_ patterns when there's more than one of each type.
There is one thing I like about always adding the 1 even if there is only a single server and that is the log naming. If I start w/ a single server process then I will get a log file w/o the 1. If I change config to running 2 server process then I will get two logs files with numbers and the log file w/o a number is abandoned.
I will take a stab at simplification.
There was a problem hiding this comment.
@ddanielr I decided to go ahead and merge this for now because we have both tested this change and did not want to test again after trying to make the code shorter.
When trying to start and stop multiple groups of scan servers using the accumulo-cluster script the groups would interfere with each other. This change fixes that by making the the scan servers use the group in the instance id like compactors do.
Ran into this while workiing on #4444 and pulled it out as a stand alone fix.