Habbix

Configuration

Copy the file config.default.yaml to ./config.yaml and edit the parameters in it.

We require access to two databases, our own “local” database and the zabbix, “remote”, database Make sure both databases are reachable. Although most commands use only the local db. The only exceptions are sync (both) and trend (remote only) commands.

Crontab


# Sync only trends and history
*/5 * * * * sh -c 'cd /home/ohtu && .cabal/bin/habbix sync -q'

# Sync new hosts too, and add them a default future
*/30 * * * * sh -c 'cd /home/ohtu && .cabal/bin/habbix sync --syncall -q'

Usage

Running habbix --help:

habbix [COMMAND] ... [OPTIONS]

Commands:
  hosts      List all hosts and groups except templates
  apps       List available "metric groups" for the Host ID
  items      List available "metrics" in the metric group App ID>
  history    Print history data for <itemid>
  trends     Print trend data for the REMOTE item
  future     List all item futures
  models     List available future models
  migratedb  Create or update the local DB schema
  sync       Synchronize remote db with local and run futures
  configure  Configure the predictions in database
  execute    Execute item_future.ID but only output the results, instead of
             modifying database
  compare    Compare predictions from knowing A to an actual history B

Common flags:
  -c --config=FILE       yaml config file (default: ./config.yaml)
  -h --human --outhuman  Human-readable JSON output
  -j --json --outjson    Bare JSON output
  -? --help              Display help message
  -V --version           Print version information
  -v --verbose           Loud verbosity
  -q --quiet             Quiet verbosity

Habbix subcommands can be broken up into categories:

Modify the database (migratedb sync configure)
Just fetch info from database (hosts apps items history future models)
Forecasting (does not modify the db in any way) (execute compare)

Run habbix subcommand --help to see additional parameters.

STDOUT is always json and by default formatted human-readable (--human). You can output bare json without newlines with (--json).

`habbix migratedb`

Creates the local database schema. (Or migrates if it exists already)

`habbix sync`

Populate or update local db with data from remote db: habbix sync.

Update the future of a specific item_future.id: habbix sync -i 2.

Careful with the populate/update outputs when running with –verbose! There is a lot of sql debug msgs.

`habbix trend`

This command uses the remote database and outputs an sql transaction that can be used to insert a data set into some other table. It outputs the transaction to stdout. Example:

$ habbix trends 23687 \       # replace number with an items.itemid in zabbix db
     --sql \                  # required. output as a transaction
     --config myconfig.yaml \ # optional. set some other config that ./config.yaml
     --asitem 1 --ashost 1    # required. itemid and hostid used in the output (only)
BEGIN;
INSERT INTO hosts VALUES (1, 'ohtu1', 0, 1, 'ohtu1');
INSERT INTO items VALUES (1, 0, 1, 'CPU $2 time', 'system.cpu.util[,idle]', 'The time the CPU has spent doing nothing.', '0');
INSERT INTO trend VALUES 
   (1, 1410501600, 94.0244, 98.8033, 99.2991),
   (1, 1410505200, 98.8481, 99.0174, 99.1656),
   (1, 1410508800, 98.7310, 99.0362, 99.2322),
   ...
COMMIT;

You can pipe the output straight to psql or to a file first.

Futures

History synchronization is done for items present in the item_future table. You need to add items to this table yourself. Note that we couple history with future: when adding an item to be synced, you must specify a future prediction model for it.

Available prediction models are stored in future_model table. future_model.name describes a binary in $PWD/forecast_models/$future_model.name, that gets executed every time a synchronization is done. The prediction program is fed a Event JSON object in its standard input, and a Result JSON object is expected as standard output.

Static linking?

Very tricky, because haskell package postgresql-libpq requires pq library which requires krb5 library, and linking (recentish) krb5 statically is unsupported and won’t work.