Hello,
we want to get a measurement on resources that still might be available in our elasticsearch cluster.
To get an idea we run Rally 0.9.:
esrally --track="http_logs" --target-hosts="node1, node2" --pipeline=benchmark-only --client-options="use_ssl:true,verify_certs:false,basic_auth_user:'superuser',basic_auth_password:'...'"
We now have difficulties to interpret the report:
Lap | Metric | Task | Value | Unit |
---|---|---|---|---|
All | Indexing time | 0 | min | |
All | Indexing throttle time | 0 | min | |
All | Merge time | 0 | min | |
All | Refresh time | 0 | min | |
All | Flush time | 0 | min | |
All | Merge throttle time | 0 | min | |
All | Total Young Gen GC | 0 | s | |
All | Total Old Gen GC | 0 | s | |
All | Min Throughput | index-append | 74604.3 | docs/s |
All | Median Throughput | index-append | 76929.9 | docs/s |
All | Max Throughput | index-append | 77214.6 | docs/s |
All | 50th percentile latency | index-append | 445.248 | ms |
All | 90th percentile latency | index-append | 693.113 | ms |
All | 99th percentile latency | index-append | 1090.24 | ms |
All | 99.9th percentile latency | index-append | 1497.43 | ms |
All | 99.99th percentile latency | index-append | 1918.44 | ms |
All | 100th percentile latency | index-append | 2313.48 | ms |
All | 50th percentile service time | index-append | 445.248 | ms |
All | 90th percentile service time | index-append | 693.113 | ms |
All | 99th percentile service time | index-append | 1090.24 | ms |
All | 99.9th percentile service time | index-append | 1497.43 | ms |
All | 99.99th percentile service time | index-append | 1918.44 | ms |
All | 100th percentile service time | index-append | 2313.48 | ms |
All | error rate | index-append | 0 | % |
All | I CUT term, hourly_agg, index-stats and node stats to match body limitation. | |||
All | Min Throughput | default | 3.33 | ops/s |
All | Median Throughput | default | 3.36 | ops/s |
All | Max Throughput | default | 3.38 | ops/s |
All | 50th percentile latency | default | 69916.2 | ms |
All | 50th percentile latency | default | 69916.2 | ms |
All | 90th percentile latency | default | 108802 | ms |
All | 99th percentile latency | default | 117133 | ms |
All | 100th percentile latency | default | 118069 | ms |
All | 50th percentile service time | default | 285.997 | ms |
All | 90th percentile service time | default | 344.239 | ms |
All | 99th percentile service time | default | 452.941 | ms |
All | 100th percentile service time | default | 547.525 | ms |
All | error rate | default | 0 | % |
All | Min Throughput | range | 1.76 | ops/s |
All | Median Throughput | range | 1.78 | ops/s |
All | Max Throughput | range | 1.79 | ops/s |
All | 50th percentile latency | range | 12738.3 | ms |
All | 90th percentile latency | range | 19092.1 | ms |
All | 99th percentile latency | range | 20829.9 | ms |
All | 100th percentile latency | range | 21039.9 | ms |
All | 50th percentile service time | range | 559.023 | ms |
All | 90th percentile service time | range | 665.189 | ms |
All | 99th percentile service time | range | 852.438 | ms |
All | 100th percentile service time | range | 1131.42 | ms |
All | error rate | range | 0 | % |
All | Min Throughput | scroll | 19.45 | ops/s |
All | Median Throughput | scroll | 20.25 | ops/s |
All | Max Throughput | scroll | 20.84 | ops/s |
All | 50th percentile latency | scroll | 228750 | ms |
All | 90th percentile latency | scroll | 313780 | ms |
All | 99th percentile latency | scroll | 332907 | ms |
All | 100th percentile latency | scroll | 335018 | ms |
All | 50th percentile service time | scroll | 1151.23 | ms |
All | 90th percentile service time | scroll | 1516.23 | ms |
All | 99th percentile service time | scroll | 2033.21 | ms |
All | 100th percentile service time | scroll | 2359.8 | ms |
All | error rate | scroll | 0 | % |
The questions are:
- Why are there no values at the top of the table for 'Indexing time', 'Indexing throttle time'... How can it be fixed?
- I can guess it but am I right to say that:
| All | Median Throughput | index-append | **76929.9** | docs/s |
means that the cluster is able to index 76929.9 documents per second? We need such information to limit new clients that send documents to index. - Is there any good measurement to get an idea on search performance (how many users can run simple searches like '*') while the elasticsearch nodes are continuously indexing documents?
- What would be the main conclusion reading this report regarding search and index performance?
Could you please help to interpret the report or point to further documentation?
Thank you very much!
Regards,
ize0j10