|
| 1 | +--- |
| 2 | +order: 1 |
| 3 | +title: Method |
| 4 | +--- |
| 5 | + |
| 6 | +# Method |
| 7 | + |
| 8 | +This document provides a detailed description of the QA process. |
| 9 | +It is intended to be used by engineers reproducing the experimental setup for future tests of Tendermint. |
| 10 | + |
| 11 | +The (first iteration of the) QA process as described [in the RELEASES.md document][releases] |
| 12 | +was applied to version v0.34.x in order to have a set of results acting as benchmarking baseline. |
| 13 | +This baseline is then compared with results obtained in later versions. |
| 14 | + |
| 15 | +Out of the testnet-based test cases described in [the releases document][releases] we focused on two of them: |
| 16 | +_200 Node Test_, and _Rotating Nodes Test_. |
| 17 | + |
| 18 | +[releases]: https://github.com/tendermint/tendermint/blob/v0.37.x/RELEASES.md#large-scale-testnets |
| 19 | + |
| 20 | +## Software Dependencies |
| 21 | + |
| 22 | +### Infrastructure Requirements to Run the Tests |
| 23 | + |
| 24 | +* An account at Digital Ocean (DO), with a high droplet limit (>202) |
| 25 | +* The machine to orchestrate the tests should have the following installed: |
| 26 | + * A clone of the [testnet repository][testnet-repo] |
| 27 | + * This repository contains all the scripts mentioned in the reminder of this section |
| 28 | + * [Digital Ocean CLI][doctl] |
| 29 | + * [Terraform CLI][Terraform] |
| 30 | + * [Ansible CLI][Ansible] |
| 31 | + |
| 32 | +[testnet-repo]: https://github.com/interchainio/tendermint-testnet |
| 33 | +[Ansible]: https://docs.ansible.com/ansible/latest/index.html |
| 34 | +[Terraform]: https://www.terraform.io/docs |
| 35 | +[doctl]: https://docs.digitalocean.com/reference/doctl/how-to/install/ |
| 36 | + |
| 37 | +### Requirements for Result Extraction |
| 38 | + |
| 39 | +* Matlab or Octave |
| 40 | +* [Prometheus][prometheus] server installed |
| 41 | +* blockstore DB of one of the full nodes in the testnet |
| 42 | +* Prometheus DB |
| 43 | + |
| 44 | +[prometheus]: https://prometheus.io/ |
| 45 | + |
| 46 | +## 200 Node Testnet |
| 47 | + |
| 48 | +### Running the test |
| 49 | + |
| 50 | +This section explains how the tests were carried out for reproducibility purposes. |
| 51 | + |
| 52 | +1. [If you haven't done it before] |
| 53 | + Follow steps 1-4 of the `README.md` at the top of the testnet repository to configure Terraform, and `doctl`. |
| 54 | +2. Copy file `testnets/testnet200.toml` onto `testnet.toml` (do NOT commit this change) |
| 55 | +3. Set the variable `VERSION_TAG` in the `Makefile` to the git hash that is to be tested. |
| 56 | +4. Follow steps 5-10 of the `README.md` to configure and start the 200 node testnet |
| 57 | + * WARNING: Do NOT forget to run `make terraform-destroy` as soon as you are done with the tests (see step 9) |
| 58 | +5. As a sanity check, connect to the Prometheus node's web interface and check the graph for the `tendermint_consensus_height` metric. |
| 59 | + All nodes should be increasing their heights. |
| 60 | +6. `ssh` into the `testnet-load-runner`, then copy script `script/200-node-loadscript.sh` and run it from the load runner node. |
| 61 | + * Before running it, you need to edit the script to provide the IP address of a full node. |
| 62 | + This node will receive all transactions from the load runner node. |
| 63 | + * This script will take about 40 mins to run |
| 64 | + * It is running 90-seconds-long experiments in a loop with different loads |
| 65 | +7. Run `make retrieve-data` to gather all relevant data from the testnet into the orchestrating machine |
| 66 | +8. Verify that the data was collected without errors |
| 67 | + * at least one blockstore DB for a Tendermint validator |
| 68 | + * the Prometheus database from the Prometheus node |
| 69 | + * for extra care, you can run `zip -T` on the `prometheus.zip` file and (one of) the `blockstore.db.zip` file(s) |
| 70 | +9. **Run `make terraform-destroy`** |
| 71 | + * Don't forget to type `yes`! Otherwise you're in trouble. |
| 72 | + |
| 73 | +### Result Extraction |
| 74 | + |
| 75 | +The method for extracting the results described here is highly manual (and exploratory) at this stage. |
| 76 | +The Core team should improve it at every iteration to increase the amount of automation. |
| 77 | + |
| 78 | +#### Steps |
| 79 | + |
| 80 | +1. Unzip the blockstore into a directory |
| 81 | +2. Extract the latency report and the raw latencies for all the experiments. Run these commands from the directory containing the blockstore |
| 82 | + * `go run github.com/tendermint/tendermint/test/loadtime/cmd/report@3ec6e424d --database-type goleveldb --data-dir ./ > results/report.txt` |
| 83 | + * `go run github.com/tendermint/tendermint/test/loadtime/cmd/report@3ec6e424d --database-type goleveldb --data-dir ./ --csv results/raw.csv` |
| 84 | +3. File `report.txt` contains an unordered list of experiments with varying concurrent connections and transaction rate |
| 85 | + * Create files `report01.txt`, `report02.txt`, `report04.txt` and, for each experiment in file `report.txt`, |
| 86 | + copy its related lines to the filename that matches the number of connections. |
| 87 | + * Sort the experiments in `report01.txt` in ascending tx rate order. Likewise for `report02.txt` and `report04.txt`. |
| 88 | +4. Generate file `report_tabbed.txt` by showing the contents `report01.txt`, `report02.txt`, `report04.txt` side by side |
| 89 | + * This effectively creates a table where rows are a particular tx rate and columns are a particular number of websocket connections. |
| 90 | +5. Extract the raw latencies from file `raw.csv` using the following bash loop. This creates a `.csv` file and a `.dat` file per experiment. |
| 91 | + The format of the `.dat` files is amenable to loading them as matrices in Octave |
| 92 | + |
| 93 | + ```bash |
| 94 | + uuids=($(cat report01.txt report02.txt report04.txt | grep '^Experiment ID: ' | awk '{ print $3 }')) |
| 95 | + c=1 |
| 96 | + for i in 01 02 04; do |
| 97 | + for j in 0025 0050 0100 0200; do |
| 98 | + echo $i $j $c "${uuids[$c]}" |
| 99 | + filename=c${i}_r${j} |
| 100 | + grep ${uuids[$c]} raw.csv > ${filename}.csv |
| 101 | + cat ${filename}.csv | tr , ' ' | awk '{ print $2, $3 }' > ${filename}.dat |
| 102 | + c=$(expr $c + 1) |
| 103 | + done |
| 104 | + done |
| 105 | + ``` |
| 106 | + |
| 107 | +6. Enter Octave |
| 108 | +7. Load all `.dat` files generated in step 5 into matrices using this Octave code snippet |
| 109 | + |
| 110 | + ```octave |
| 111 | + conns = { "01"; "02"; "04" }; |
| 112 | + rates = { "0025"; "0050"; "0100"; "0200" }; |
| 113 | + for i = 1:length(conns) |
| 114 | + for j = 1:length(rates) |
| 115 | + filename = strcat("c", conns{i}, "_r", rates{j}, ".dat"); |
| 116 | + load("-ascii", filename); |
| 117 | + endfor |
| 118 | + endfor |
| 119 | + ``` |
| 120 | +
|
| 121 | +8. Set variable release to the current release undergoing QA |
| 122 | +
|
| 123 | + ```octave |
| 124 | + release = "v0.34.x"; |
| 125 | + ``` |
| 126 | +
|
| 127 | +9. Generate a plot with all (or some) experiments, where the X axis is the experiment time, |
| 128 | + and the y axis is the latency of transactions. |
| 129 | + The following snippet plots all experiments. |
| 130 | +
|
| 131 | + ```octave |
| 132 | + legends = {}; |
| 133 | + hold off; |
| 134 | + for i = 1:length(conns) |
| 135 | + for j = 1:length(rates) |
| 136 | + data_name = strcat("c", conns{i}, "_r", rates{j}); |
| 137 | + l = strcat("c=", conns{i}, " r=", rates{j}); |
| 138 | + m = eval(data_name); plot((m(:,1) - min(m(:,1))) / 1e+9, m(:,2) / 1e+9, "."); |
| 139 | + hold on; |
| 140 | + legends(1, end+1) = l; |
| 141 | + endfor |
| 142 | + endfor |
| 143 | + legend(legends, "location", "northeastoutside"); |
| 144 | + xlabel("experiment time (s)"); |
| 145 | + ylabel("latency (s)"); |
| 146 | + t = sprintf("200-node testnet - %s", release); |
| 147 | + title(t); |
| 148 | + ``` |
| 149 | +
|
| 150 | +10. Consider adjusting the axis, in case you want to compare your results to the baseline, for instance |
| 151 | +
|
| 152 | + ```octave |
| 153 | + axis([0, 100, 0, 30], "tic"); |
| 154 | + ``` |
| 155 | +
|
| 156 | +11. Use Octave's GUI menu to save the plot (e.g. as `.png`) |
| 157 | +
|
| 158 | +12. Repeat steps 9 and 10 to obtain as many plots as deemed necessary. |
| 159 | +
|
| 160 | +13. To generate a latency vs throughput plot, using the raw CSV file generated |
| 161 | + in step 2, follow the instructions for the [`latency_throughput.py`] script. |
| 162 | +
|
| 163 | +[`latency_throughput.py`]: ../../scripts/qa/reporting/README.md |
| 164 | +
|
| 165 | +#### Extracting Prometheus Metrics |
| 166 | +
|
| 167 | +1. Stop the prometheus server if it is running as a service (e.g. a `systemd` unit). |
| 168 | +2. Unzip the prometheus database retrieved from the testnet, and move it to replace the |
| 169 | + local prometheus database. |
| 170 | +3. Start the prometheus server and make sure no error logs appear at start up. |
| 171 | +4. Introduce the metrics you want to gather or plot. |
| 172 | +
|
| 173 | +## Rotating Node Testnet |
| 174 | +
|
| 175 | +### Running the test |
| 176 | +
|
| 177 | +This section explains how the tests were carried out for reproducibility purposes. |
| 178 | +
|
| 179 | +1. [If you haven't done it before] |
| 180 | + Follow steps 1-4 of the `README.md` at the top of the testnet repository to configure Terraform, and `doctl`. |
| 181 | +2. Copy file `testnet_rotating.toml` onto `testnet.toml` (do NOT commit this change) |
| 182 | +3. Set variable `VERSION_TAG` to the git hash that is to be tested. |
| 183 | +4. Run `make terraform-apply EPHEMERAL_SIZE=25` |
| 184 | + * WARNING: Do NOT forget to run `make terraform-destroy` as soon as you are done with the tests |
| 185 | +5. Follow steps 6-10 of the `README.md` to configure and start the "stable" part of the rotating node testnet |
| 186 | +6. As a sanity check, connect to the Prometheus node's web interface and check the graph for the `tendermint_consensus_height` metric. |
| 187 | + All nodes should be increasing their heights. |
| 188 | +7. On a different shell, |
| 189 | + * run `make runload ROTATE_CONNECTIONS=X ROTATE_TX_RATE=Y` |
| 190 | + * `X` and `Y` should reflect a load below the saturation point (see, e.g., |
| 191 | + [this paragraph](./v034/README.md#finding-the-saturation-point) for further info) |
| 192 | +8. Run `make rotate` to start the script that creates the ephemeral nodes, and kills them when they are caught up. |
| 193 | + * WARNING: If you run this command from your laptop, the laptop needs to be up and connected for full length |
| 194 | + of the experiment. |
| 195 | +9. When the height of the chain reaches 3000, stop the `make rotate` script |
| 196 | +10. When the rotate script has made two iterations (i.e., all ephemeral nodes have caught up twice) |
| 197 | + after height 3000 was reached, stop `make rotate` |
| 198 | +11. Run `make retrieve-data` to gather all relevant data from the testnet into the orchestrating machine |
| 199 | +12. Verify that the data was collected without errors |
| 200 | + * at least one blockstore DB for a Tendermint validator |
| 201 | + * the Prometheus database from the Prometheus node |
| 202 | + * for extra care, you can run `zip -T` on the `prometheus.zip` file and (one of) the `blockstore.db.zip` file(s) |
| 203 | +13. **Run `make terraform-destroy`** |
| 204 | +
|
| 205 | +Steps 8 to 10 are highly manual at the moment and will be improved in next iterations. |
| 206 | +
|
| 207 | +### Result Extraction |
| 208 | +
|
| 209 | +In order to obtain a latency plot, follow the instructions above for the 200 node experiment, but: |
| 210 | +
|
| 211 | +* The `results.txt` file contains only one experiment |
| 212 | +* Therefore, no need for any `for` loops |
| 213 | +
|
| 214 | +As for prometheus, the same method as for the 200 node experiment can be applied. |
![mergify[bot]](https://avatars.githubusercontent.com/in/10562?v=4&size=40){kind=avatar}
0 commit comments