A script written and used by Voxer to check Riak health on SmartOS
Note: Only tested on SmartOS
You can run check-riak directly out of the git repository. To install
it globally on the system, run:
make install
and ensure /opt/local/bin is in your PATH
To remove check-riak, run
make uninstall
usage: check-riak [ options ] [ check ]
run this script with no options to get an overview of riak health
on a system.
give an optional parameter to only check one portion of the system.
examples
check-riak # full system health overview
check-riak compaction # only check for compaction errors
check-riak -n compaction # check for compaction, output suitable for nagios
options
-a run all checks
-c <config> the riak config file to use, defaults to /opt/local/etc/riak/app.config
-d <dir> the database directory, defaults to /var/db/riak/leveldb
-h print this message and exit
-H <host> the riak host, defaults to localhost
-n less pretty output suitable for a nagios check
-p <port> the riak port, defaults to 8098
-s <svc> the service fmri to use for SMF, defaults to riak
-t <seconds> the timeout, in seconds, when curl'ing riak, defaults to 10
-W <bytes> the warning threshold for rss in bytes, defaults to 50000000000 (50g)
-C <bytes> the critical threshold for rss in bytes, defaults to 60000000000 (60g)
optional arguments
compaction check for compaction errors
config show the first line of the riak config
oktostart check if riak is ok to start
ping check riak ping using /ping
process parse ps for the running process
profile profile beam.smp with DTrace
riakping check riak ping locally
rss check riak rss
service show svcs output for riak
singleton see if this machine is a single cluster
stats check riak stats (riak_kv_vnodes_running)
system check riak system health
Run check-riak with no arguments to see health of a node
$ sudo check-riak
Riak Cluster
%% Test Cluster (1.3.0)
Checking Service
STATE STIME FMRI
online 13:24:29 svc:/application/riak:default
Checking Process
riak 58798 58796 0 13:24:22 pts/8 0:08 /opt/local/lib/riak/erts-5.9.1/bin/beam.smp -K true -A 64 -W w -- -root /opt/lo
System Health
pid 58798: 65,672K rss (279 open files)
Checking Riak Ping
pong
Checking Riak Stats Ping
OK
Checking Singleton
ok: 16 running of 64 total (25%)
Checking Compaction Errors
-------- Compaction errors found! --------
The following steps will resolve them
** stop the process before running!! **
# Run the following command(s) one-by-one
# (safe to run multiple commands at the same time to saturate disks)
/opt/local/lib/riak/erts-5.9.1/bin/escript "./lib/lvlfix.erl" "/var/db/riak/leveldb/356811923176489970264571492362373784095686656000"
There's a lot of output, so let's go through it section by section.
This section is very simple, it prints out the first line (head -1) of the
Riak app.config file as well as the Riak version number.
At Voxer, we have multiple Riak clusters and we put the name of the cluster as a comment on the first line of the config, so this helps us identify which cluster we are currently checking.
This will check SMF on machines with svcs to show the current state
of the service.
If your machine lacks SMF and the svcs binary, this check will be skipped
Just a simple parsing of ps for the riak beam process
This is useful for situations where SMF shows the process as disabled or in maintenance mode, but the process is still up and running.
Shows the PID, RSS, and # of open files for Riak.
This is useful in situations where SMF shows the process as enabled, the
beam process is found in ps, but the /proc structure cannot be accessed.
This is usually a sign that Riak has crashed and is writing a core file.
This is the output of riak ping, it should say pong
This is the output of a GET request to riak at /ping, it should say OK
This checks the number of running VNodes for a single riak process, and compares it
to the ring_creation_size found in the app.config file.
If the number of VNodes running is greater than or equal to the
ring_creation_size, it is considered critical. At Voxer, we don't have any
singleton clusters; all of our Riak clusters have more than 1 node in them. If
this check ever becomes critical, it means a node is running Riak, but is not
part of cluster. If it remains in the load-balancer it could potentially
result in data corruption or data loss.
In order to retrieve the number of VNodes running, the stats interface /stats
is curl'd and jsontool is used to extract the metric
riak_kv_vnodes_running.
This check looks for compaction errors for the data on the filesystem.
If there are any compaction errors found, this tool will print the steps required to resolve the errors.
In the above example, you can see that every check has passed except the check for Compaction Errors. Let's follow the directions to resolve the compaction errors.
$ sudo svcadm disable riak
$ sudo /opt/local/lib/riak/erts-5.9.1/bin/escript "./lib/lvlfix.erl" "/var/db/riak/leveldb/356811923176489970264571492362373784095686656000"
["356811923176489970264571492362373784095686656000"]
Compactions
Level Files Size(MB) Time(sec) Read(MB) Write(MB)
--------------------------------------------------
Now that the compaction errors have been resolved, let's make sure Riak is OK
to start, running check-riak we'll see different output as the service
is not currently running.
$ sudo check-riak
Riak Cluster
%% Test Cluster
Checking Service
STATE STIME FMRI
disabled 15:05:33 svc:/application/riak:default
Checking Process
not running
Checking OK To Start
All Partitions compacted, OK to start.
Checking Compaction Errors
ok: no compaction errors found
No compaction errors, and Riak is ready to start.
Some of the checks support outputting to a nagios-friendly format with the -n option.
$ sudo check-riak -n rss
ok: 65,664K rss
$ echo $?
0
$ sudo check-riak -n ping
OK
$ echo $?
0
$ sudo check-riak -n singleton
ok: 16 running of 64 total (25%)
$ echo $?
0
$ sudo check-riak -n compation
ok: no compaction errors found
$ echo $?
0
- This script should work on other operating systems, though it hasn't been tested by us
- This script assumes the first line of your
app.configis meaningful - jsontool is required for some of the checks to work; SmartOS has this installed by default
- Escalated privileges are required for certain actions, though
sudoitself isn't necessarily required - Logic to find compaction errors found here https://gist.github.com/gburd/b88aee6da7fee81dc036
LICENSE - "MIT License" Copyright (c) 2013-2014 Voxer LLC. All rights reserved.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.