hostbbq-monitoring/daemon/src/collector.class.ts

import fs, { PathLike } from 'fs';
import fsp from 'fs/promises';
import moment from 'moment';
import { exec } from 'node-utils/shell';
import path from 'path';

import { Logger } from '../../common/util/logger.class';
import { HttpStatusException } from './lib/http-status.exception';

const DATA_DIR = process.env.DATA_DIR;
const DATA_BUFFER_FILE = path.resolve(DATA_DIR, 'buffer.csv');
const DATA_BUFFER_REMAINS = path.resolve(DATA_DIR, 'remains.csv');
const DATA_REDUCED_FILE = path.resolve(DATA_DIR, 'reduced.csv');
const TIMESTAMP_FORMAT = `YYYY-MM-DD[T]HH:mm:ss.SSSZZ`;
const REDUCE_INTERVAL_MINUTES = 5;
const REDUCE_GROUP_MINUTES = 1;

const MONITOR_MOUNTS = !!process.env.MONITOR_MOUNTS ? process.env.MONITOR_MOUNTS.split(':') : [];

Logger.info('[INFO] Monitoring Drives:', MONITOR_MOUNTS);

const CSV_COLS = {
  buffer: {
    time: 0,
    cpu: 1,
    ram: 2
  },
  reduced: {
    time: 0,
    cpu: {
      avg: 1,
      peak: 2
    },
    ram: {
      avg: 3,
      peak: 4,
      max: 5
    }
  }
};

export class Collector {
  private intervalHdl: NodeJS.Timer;

  constructor() {
    (async () => {
      try {
        if (!fs.existsSync(DATA_DIR)) {
          Logger.info('[INFO] DATA_DIR', DATA_DIR, 'does not exist - creating now ...');
          await fsp.mkdir(DATA_DIR);
        }

        this.startLoop();
      } catch (err) {
        Logger.error('[FATAL]', err);
        Logger.error('[EXITING]');
        process.exit(1);
      }
    })();
  }

  public startLoop() {
    this.intervalHdl = setInterval(this.loop.bind(this), 500);
  }

  private async loop() {
    try {
      const now = moment();
      const time = now.format(TIMESTAMP_FORMAT);
      const cpu = (await exec(`./cpu.sh`)).trim();
      const ram = (await exec(`./ram.sh`)).trim();

      const hdd: string[] = [];
      for (const mount of MONITOR_MOUNTS) {
        try {
          const stats = (await exec(`./hdd.sh "${mount}"`)).trim();
          if (stats?.length) hdd.push(`${mount} ${stats}`);
        } catch (err) {
          Logger.warn('[WARN] Error while getting space usage of mount', mount, ':', err);
        }
      }

      const data = `${time};${cpu};${ram}${hdd.length ? `;${hdd.join(';')}` : ''}\n`;

      // Time to reduce buffer?
      const firstBufferTime = await this.getFirstBufferTime();
      if (moment.duration(now.diff(firstBufferTime)).abs().asMinutes() >= REDUCE_INTERVAL_MINUTES) {
        try {
          const tmpFile = await this.createTmpFile();
          process.nextTick(() => this.reduceData(tmpFile));
        } catch (err) {
          Logger.error('[ERROR] Creating Temp File for Reducing Data failed:', err);
        }
      }

      await fsp.appendFile(DATA_BUFFER_FILE, data);
    } catch (err) {
      Logger.error(err);
    }
  }

  private async getFirstBufferTime() {
    let dataFile = DATA_BUFFER_FILE;
    if (fs.existsSync(DATA_BUFFER_REMAINS)) dataFile = DATA_BUFFER_REMAINS;
    if (!fs.existsSync(dataFile)) return moment();

    const firstLine = await this.readFirstBufferLine(dataFile);
    const timestamp = firstLine.split(';')[CSV_COLS.buffer.time];
    return moment(timestamp, TIMESTAMP_FORMAT);
  }

  private readFirstBufferLine = (dataFile: PathLike) =>
    new Promise<string>((resolve, reject) => {
      const stream = fs.createReadStream(dataFile, { encoding: 'utf-8' });
      const chunks: string[] = [];
      stream
        .on('data', buf => {
          let chunk: string;
          if (buf instanceof Buffer) chunk = buf.toString('utf-8');
          else chunk = buf;

          const lfIdx = chunk.indexOf('\n');
          if (lfIdx >= 0) {
            chunks.push(chunk.substring(0, lfIdx));
            stream.close();
          } else {
            chunks.push(chunk);
          }
        })
        .on('close', () => resolve(chunks.join('')))
        .on('error', reject);
    });

  private async readDataFileCSV(dataFile: PathLike): Promise<string[][]> {
    return (await fsp.readFile(dataFile, { encoding: 'utf-8' }))
      .split(/\r?\n/g)
      .filter(l => !!l)
      .map(line => line.split(';'));
  }

  private async createTmpFile() {
    const tmpFilename = `buffer.tmp_${moment().format('YYYYMMDDHHmmssSSS')}.csv`;
    await fsp.rename(DATA_BUFFER_FILE, path.resolve(DATA_DIR, tmpFilename));
    return tmpFilename;
  }

  private async reduceData(tmpFilename: string) {
    const tmpFilepath = path.resolve(DATA_DIR, tmpFilename);
    Logger.info('[INFO] Reducing data in', tmpFilepath);
    try {
      const lines: string[][] = [];
      if (fs.existsSync(DATA_BUFFER_REMAINS)) {
        lines.push(...(await this.readDataFileCSV(DATA_BUFFER_REMAINS)));
      }
      lines.push(...(await this.readDataFileCSV(tmpFilepath)));

      const reduced: Array<ReducedData> = [];
      let valueBuffer: Array<BufferedData> = [];
      do {
        const line = lines.shift();
        const data = this.parseBufferedData(line);
        Logger.debug('[DEBUG] BufferedData:', JSON.stringify(data));
        valueBuffer.push(data);

        if (valueBuffer.length <= 1) {
          // Need at least 2 datasets to check for time diff and eventually reduce to avg/max.
          // skip to next data line
          continue;
        }

        const firstTime = moment(valueBuffer[0].time);
        const currentTime = moment(data.time);
        if (moment.duration(currentTime.diff(firstTime)).abs().asMinutes() >= REDUCE_GROUP_MINUTES) {
          type IntermediateValues = { sum: number; peak: number; max: number };
          type IntermediateDriveData = { [mount: string]: IntermediateValues };
          type IntermediateSums = { ram: IntermediateValues; cpu: IntermediateValues; hdd?: IntermediateDriveData; count: number };
          const { cpu, ram, count, hdd } = valueBuffer.reduce(
            (res, cur) => {
              res.count++;
              res.cpu.sum += cur.cpu;
              res.cpu.peak = Math.max(res.cpu.peak, cur.cpu);
              res.ram.sum += cur.ram.used;
              res.ram.peak = Math.max(res.ram.peak, cur.ram.used);
              res.ram.max = cur.ram.max;

              if (cur.hdd && Object.keys(cur.hdd).length) {
                const hdd_sums = res.hdd ?? {};
                res.hdd = Object.keys(cur.hdd).reduce((res_hdd, mount) => {
                  if (!res_hdd[mount]) {
                    res_hdd[mount] = { sum: 0, peak: 0, max: 0 };
                  }
                  res_hdd[mount].sum += cur.hdd[mount].used;
                  res_hdd[mount].peak = Math.max(res_hdd[mount].peak, cur.hdd[mount].used);
                  res_hdd[mount].max = cur.hdd[mount].max;
                  return res_hdd;
                }, hdd_sums);
              }

              return res;
            },
            { ram: { sum: 0, peak: 0, max: 0 }, cpu: { sum: 0, peak: 0 }, count: 0 } as IntermediateSums
          );

          reduced.push({
            time: data.time,
            cpu: {
              avg: cpu.sum / count,
              peak: cpu.peak
            },
            ram: {
              avg: ram.sum / count,
              peak: ram.peak,
              max: ram.max
            },
            hdd: hdd
              ? Object.keys(hdd).reduce((res, mount) => {
                  res[mount] = {
                    avg: hdd[mount].sum / count,
                    peak: hdd[mount].peak,
                    max: hdd[mount].max
                  };
                  return res;
                }, {} as ReducedDriveData)
              : undefined
          });

          Logger.debug('[DEBUG] ReducedData:', JSON.stringify(reduced[reduced.length - 1]));

          valueBuffer = [];
        }
      } while (lines.length > 0);

      if (valueBuffer.length > 0) {
        // overwrite remains.csv with valueBuffer
        await fsp.writeFile(DATA_BUFFER_REMAINS, valueBuffer.map(this.serializeBufferedDataCSV.bind(this)).join('\n') + '\n', { encoding: 'utf-8' });
      } else {
        // delete remains.csv if exists
        if (fs.existsSync(DATA_BUFFER_REMAINS)) await fsp.unlink(DATA_BUFFER_REMAINS);
      }

      if (reduced.length > 0) {
        // append reduced data to reduced.csv
        await fsp.appendFile(DATA_REDUCED_FILE, reduced.map(this.serializeReducedDataCSV.bind(this)).join('\n') + '\n', { encoding: 'utf-8' });
      }

      // Delete tmpFile
      await fsp.unlink(tmpFilepath);
    } catch (err) {
      Logger.error(`[ERROR] Reducing Data of tmpFile ${tmpFilepath} failed:`, err);
    }
  }

  private parseBufferedData(line: string[]): BufferedData {
    // TIMESTAMP
    const time = moment(line[CSV_COLS.buffer.time], TIMESTAMP_FORMAT).toDate();

    // CPU
    const cpu = Number(line[CSV_COLS.buffer.cpu]);

    // RAM
    let [stats, unit] = line[CSV_COLS.buffer.ram].split(' ');
    const [used, max] = stats.split('/');
    const factor = this.parseByteUnit(unit);

    const lastCol = CSV_COLS.buffer.ram;

    // HDD (?)
    let hdd: BufferedDriveData;
    if (MONITOR_MOUNTS.length && line.length > lastCol + 1) {
      for (let i = 1; i <= MONITOR_MOUNTS.length; i++) {
        if (lastCol + i > line.length - 1) break;

        const data = line[lastCol + i];
        const [mount, stats] = data.split(' ');
        const [used, max] = stats.split('/');

        if (!hdd) hdd = {};
        hdd[mount] = {
          used: Number(used),
          max: Number(max)
        };
      }
    }

    return {
      time,
      cpu,
      ram: {
        used: Number(used) * factor,
        max: Number(max) * factor
      },
      hdd
    };
  }

  private parseReducedData(line: string[]): ReducedData {
    const lastCol = CSV_COLS.reduced.ram.max;

    // HDD (?)
    let hdd: ReducedDriveData;
    if (MONITOR_MOUNTS.length && line.length > lastCol + 1) {
      hdd = {};
      for (let i = 1; lastCol + i + 3 < line.length; i += 4) {
        hdd[line[lastCol + i]] = {
          avg: Number(line[lastCol + i + 1]),
          peak: Number(line[lastCol + i + 2]),
          max: Number(line[lastCol + i + 3])
        };
      }
    }

    return {
      time: moment(line[CSV_COLS.reduced.time], TIMESTAMP_FORMAT).toDate(),
      cpu: {
        avg: Number(line[CSV_COLS.reduced.cpu.avg]),
        peak: Number(line[CSV_COLS.reduced.cpu.peak])
      },
      ram: {
        avg: Number(line[CSV_COLS.reduced.ram.avg]),
        peak: Number(line[CSV_COLS.reduced.ram.peak]),
        max: Number(line[CSV_COLS.reduced.ram.max])
      },
      hdd
    };
  }

  private byteFactors: { [unit: string]: number } = {
    '': 1,
    K: 1024,
    M: 1024 * 1024,
    G: 1024 * 1024 * 1024,
    T: 1024 * 1024 * 1024 * 1024,
    P: 1024 * 1024 * 1024 * 1024 * 1024
  };

  private parseByteUnit(unit: string): number {
    const m = /^([KMGTP])?i?B$/.exec(unit);
    if (!m) throw new Error(`Failed to parse byte size unit '${unit}'`);

    return this.byteFactors[m[1]];
  }

  private serializeBufferedDataCSV(data: BufferedData) {
    return [
      moment(data.time).format(TIMESTAMP_FORMAT),
      data.cpu,
      `${(data.ram.used / this.byteFactors['M']).toFixed(2)}/${(data.ram.max / this.byteFactors['M']).toFixed(2)} MiB`,
      ...(data.hdd ? Object.keys(data.hdd).map(mount => `${mount} ${data.hdd[mount].used}/${data.hdd[mount].max}`) : [])
    ].join(';');
  }

  private serializeReducedDataCSV(data: ReducedData) {
    return [
      moment(data.time).format(TIMESTAMP_FORMAT),
      data.cpu.avg.toFixed(2),
      data.cpu.peak.toFixed(2),
      data.ram.avg.toFixed(2),
      data.ram.peak.toFixed(2),
      data.ram.max.toFixed(2),
      ...(data.hdd
        ? Object.keys(data.hdd).reduce((res, mount) => {
            res.push(mount, data.hdd[mount].avg.toFixed(2), data.hdd[mount].peak.toFixed(2), data.hdd[mount].max.toFixed(2));
            return res;
          }, [])
        : [])
    ].join(';');
  }

  public get trx() {
    return this._trx;
  }

  private _trx: {
    file?: PathLike;
    start: () => Promise<number>;
    read: () => Promise<Array<ReducedData>>;
    rollback: (hdl: number) => Promise<void>;
    commit: (hdl: number) => Promise<void>;
  } = {
    start: async () => {
      if (this.trx.file) {
        Logger.warn(`[WARN] Old transaction file found - rolling back now before starting new transaction ...`);
        const hdl = Number(/trx_(\d+)\.csv/.exec(this.trx.file as string)[1]);
        await this.trx.rollback(hdl);
        Logger.warn(`[WARN] Transaction rollback succeeded.`);
      }

      if (!fs.existsSync(DATA_REDUCED_FILE)) {
        // NO DATA
        return null;
      }

      const hdl = moment().unix();
      this.trx.file = path.resolve(DATA_DIR, `reduced.trx_${hdl.toFixed(0)}.csv`);
      await fsp.rename(DATA_REDUCED_FILE, this.trx.file);
      return hdl;
    },
    read: async () => {
      if (!this.trx.file) throw new Error('No transaction opened');

      const data = await this.readDataFileCSV(this.trx.file);
      return data.map(this.parseReducedData.bind(this));
    },
    rollback: async (hdl: number) => {
      if (this.trx.file) {
        const filename = path.resolve(DATA_DIR, `reduced.trx_${hdl.toFixed(0)}.csv`);
        if (filename !== this.trx.file) throw new HttpStatusException(`Transaction #${hdl} not found`, 404);

        if (fs.existsSync(this.trx.file)) {
          let tmpFile: string;
          if (fs.existsSync(DATA_REDUCED_FILE)) {
            tmpFile = `reduced.tmp_${moment().unix().toFixed(0)}.csv`;
            await fsp.rename(DATA_REDUCED_FILE, tmpFile);
          }
          await fsp.rename(this.trx.file, DATA_REDUCED_FILE);
          if (tmpFile) {
            await exec(`cat "${tmpFile}" >> "${DATA_REDUCED_FILE}"`);
            await fsp.unlink(tmpFile);
          }
        }
        this.trx.file = undefined;
      }
    },
    commit: async (hdl: number) => {
      if (this.trx.file) {
        const filename = path.resolve(DATA_DIR, `reduced.trx_${hdl.toFixed(0)}.csv`);
        if (filename !== this.trx.file) throw new HttpStatusException(`Transaction #${hdl} not found`, 404);

        if (fs.existsSync(this.trx.file)) {
          await fsp.unlink(this.trx.file);
        }
        this.trx.file = undefined;
      }
    }
  };

  public stopLoop() {
    clearInterval(this.intervalHdl);
  }
}