| 1 |
# Cluster Test |
| 2 |
|
| 3 |
This directory contains a program you can use to test a cluster. |
| 4 |
|
| 5 |
Here's how: |
| 6 |
|
| 7 |
First, install a cluster of Elasticsearch nodes. You can install them on |
| 8 |
different computers, or start several nodes on a single machine. |
| 9 |
|
| 10 |
Build cluster-test by `go build cluster-test.go` (or build with `make`). |
| 11 |
|
| 12 |
Run `./cluster-test -h` to get a list of flags: |
| 13 |
|
| 14 |
```sh |
| 15 |
$ ./cluster-test -h |
| 16 |
Usage of ./cluster-test: |
| 17 |
-errorlog="": error log file |
| 18 |
-healthcheck=true: enable or disable healthchecks |
| 19 |
-healthchecker=1m0s: healthcheck interval |
| 20 |
-index="twitter": name of ES index to use |
| 21 |
-infolog="": info log file |
| 22 |
-n=5: number of goroutines that run searches |
| 23 |
-nodes="": comma-separated list of ES URLs (e.g. 'http://192.168.2.10:9200,http://192.168.2.11:9200') |
| 24 |
-retries=0: number of retries |
| 25 |
-sniff=true: enable or disable sniffer |
| 26 |
-sniffer=15m0s: sniffer interval |
| 27 |
-tracelog="": trace log file |
| 28 |
``` |
| 29 |
|
| 30 |
Example: |
| 31 |
|
| 32 |
```sh |
| 33 |
$ ./cluster-test -nodes=http://127.0.0.1:9200,http://127.0.0.1:9201,http://127.0.0.1:9202 -n=5 -index=twitter -retries=5 -sniff=true -sniffer=10s -healthcheck=true -healthchecker=5s -errorlog=error.log |
| 34 |
``` |
| 35 |
|
| 36 |
The above example will create an index and start some search jobs on the |
| 37 |
cluster defined by http://127.0.0.1:9200, http://127.0.0.1:9201, |
| 38 |
and http://127.0.0.1:9202. |
| 39 |
|
| 40 |
* It will create an index called `twitter` on the cluster (`-index=twitter`) |
| 41 |
* It will run 5 search jobs in parallel (`-n=5`). |
| 42 |
* It will retry failed requests 5 times (`-retries=5`). |
| 43 |
* It will sniff the cluster periodically (`-sniff=true`). |
| 44 |
* It will sniff the cluster every 10 seconds (`-sniffer=10s`). |
| 45 |
* It will perform health checks periodically (`-healthcheck=true`). |
| 46 |
* It will perform health checks on the nodes every 5 seconds (`-healthchecker=5s`). |
| 47 |
* It will write an error log file (`-errorlog=error.log`). |
| 48 |
|
| 49 |
If you want to test Elastic with nodes going up and down, you can use a |
| 50 |
chaos monkey script like this and run it on the nodes of your cluster: |
| 51 |
|
| 52 |
```sh |
| 53 |
#!/bin/bash |
| 54 |
while true |
| 55 |
do |
| 56 |
echo "Starting ES node" |
| 57 |
elasticsearch -d -Xmx4g -Xms1g -Des.config=elasticsearch.yml -p es.pid |
| 58 |
sleep `jot -r 1 10 300` # wait for 10-300s |
| 59 |
echo "Stopping ES node" |
| 60 |
kill -TERM `cat es.pid` |
| 61 |
sleep `jot -r 1 10 60` # wait for 10-60s |
| 62 |
done |
| 63 |
``` |