[MDEV-6432] /etc/init.d/garb requires nc and produces an nc error if no cluster nodes are up Created: 2014-07-09  Updated: 2015-03-12  Resolved: 2015-03-12

Status: Closed
Project: MariaDB Server
Component/s: Galera Arbitrator garbd
Affects Version/s: 10.0.12-galera
Fix Version/s: 5.5.42-galera, 10.0.17-galera

Type: Bug Priority: Minor
Reporter: Richard Lloyd Assignee: Nirbhay Choubey (Inactive)
Resolution: Fixed Votes: 0
Labels: galera, garb
Environment:

64-bit CentOS 6.5
galera-25.3.5-1.rhel6.x86_64 RPM



 Description   

I installed the galera 25.3.5 RPM from the http://yum.mariadb.org/10.0/centos6-amd64 repo and after configuring /etc/default/garb, I then ran "/etc/init.d/garb start" without any of the configured cluster nodes being up.

First issue is that line 122 reads:

nc -z $HOST $PORT >/dev/null && break

and yet a standard CentOS 6.5 server install doesn't include the "nc" command. I had to "yum install nc" to fix this, but the RPM should obviously have nc as a dependency.

Second problem is that the code in that loop (lines 118-123) will test each node and then set $ADDRESS = 0 (and hence $HOST =0 and $PORT = 0) if none of the ${GALERA_NODES} nodes are running. Unfortunately, this produces a port error message from CentOS 6.5's nc command because "nc -z 0 0" is, of course, not correct.

The fix for this is probably to only run the code if $ADDRESS != 0 e.g.:

for ADDRESS in ${GALERA_NODES} 0; do
if [ "$ADDRESS" != "0"]
then
HOST=$(echo $ADDRESS | cut -d \: -f 1 )
PORT=$(echo $ADDRESS | cut -d \: -f 2 )
PORT=${PORT:-$GALERA_PORT}
nc -z $HOST $PORT >/dev/null && break
fi
done



 Comments   
Comment by Richard Lloyd [ 2014-07-09 ]

I knew I'd miss some info and get something wrong too...

It is, of course, /etc/sysconfig/garb that I edited on CentOS 6.5, not /etc/default/garb. My setup involved a 2-node cluster with a third node just running garbd, so an example line in the garb conf (IPs changed to protect the innocent ) would be:

GALERA_NODES="192.168.0.1:4567 192.168.0.2:4567"

with, say, the third node being 192.168.0.3. Also note that when the cluster is up and running, "/etc/init.d/garb start" runs successfully with no errors.

Comment by Nirbhay Choubey (Inactive) [ 2014-07-12 ]

Thank you for reporting the bug and suggestions.
I have opened a bug with galera : https://bugs.launchpad.net/galera/+bug/1341028

Comment by Nirbhay Choubey (Inactive) [ 2015-03-12 ]

Fixed in galera-25.3.9.

Generated at Thu Feb 08 07:11:51 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.