Graham Eddy

weewx Server (multiple instances)

Install, configure and manage multiple instances of weewx using systemd.

2022-03-26 Raspbian buster

This is a guide to setting up multiple instances of weewx on a single Debian server using systemd. The reader is expected to be linux literate.

More accurately, it is a cookbook on how I configure my host to run separate instances of weewx to:

• collect from a Vantage Pro2;
• collect from a GW1000; and
• collect some other inputs, aggregate, and generate consolidated report.

There are several additional requirements that I implement that influence my implementation:

• weewx instances run as user weewx not as root;
• generated files are placed in ramdisk; and
• no swapping – totally RAM-resident.

Installation

tmpfs /home/weewx/public_html tmpfs rw,nodev,nosuid,noatime,noexec,uid=weewx,gid=weewx,mode=0775,size=10M

• Verify /home/weewx/public_html is mounted correctly.

• Verify no space allocated to Swap.

• Verify weewx.conf has been moved aside. Now weewx commands must specify the config file using the --config= option. Good practice: the default will fail.

• Next login by user weewx will pick up environment changes.

Base configuration to support weewx instances

Although all instances run from the same codebase, we keep individual state in files named after the instance inst:

~weewx/inst.conf

instance configuration, based upon saved weewx.conf-dist

~weewx/archive/inst.sdb

instance database

~weewx/skins/inst/

instance skin template files, based upon provided Seasons skin for the purpose of this document

~weewx/public_html/inst/

instance report files generated at run-time

/var/log/weewx/inst.log

instance log

/run/weewx/inst.pid

instance PID file

Note that we will designate one weewx instance as the producer of the official, consolidated report, but the other instances will produce smaller reports more specific to them.

For the systemd services, think of weewx@ as a class and weewx@foo as an instance of that class. We install file weewx@.service to create the class as part of our systemd configuration; we will create instances later.

# systemd weewx (unit instance)
#   pid:                        /run/weewx/%i.pid
#   log-label (programname):    weewx-%i
#   conf:                       /home/weewx/%i.conf

[Unit]
Description=weewx weather system (%I)
Requires=time-sync.target
After=time-sync.target
RequiresMountsFor=/home

[Service]
ExecStart=/home/weewx/bin/weewxd --daemon --pidfile=/run/weewx/%i.pid --log-label=weewx-%i /home/weewx/%i.conf
ExecReload=/bin/kill -HUP $MAINPID
Type=forking
PIDFile=/run/weewx/%i.pid
User=weewx
Group=weewx
# workaround to stop SIGKILL sent instantly (systemd bug)
SendSIGKILL=no

[Install]
WantedBy=multi-user.target
# no default instance
#DefaultInstance=

#type path            mode user  group age arg
d     /run/weewx      0755 weewx weewx -

# rsyslog weewx

# weewx default
:programname,startswith,"weewx" /var/log/weewx/weewx.log
:programname,startswith,"weewx" stop

# weewx utilities
:programname,startswith,"wee_" /var/log/weewx/weewx.log
:programname,startswith,"wee_" stop

# logrotate weewx

/var/log/weewx/weewx.log
{
  create 640 weewx adm
  daily
  missingok
  rotate 7
  compress
  delaycompress
  notifempty
  maxsize=10M
  sharedscripts
  postrotate
    /usr/bin/killall -HUP rsyslogd
  endscript
}

First weewx instance: vp2 collector

skin

    [[vp2Report]]             # optional for collector
        # skin for the vp2 instance
        skin = vp2
        enable = true         # enable vp2 custom report
        HTML_ROOT = public_html/vp2  # subdir

    [[SeasonsReport]]
        # The SeasonsReport uses the 'Seasons' skin, which contains the
        # images, templates and plots for the report.
        skin = Seasons
        enable = false        # disable standard report

• Update the skin files for the reports specific to vp2. This is outside the scope of this document.

database

    [[archive_sqlite]]
        database_name = vp2.sdb
        database_type = SQLite

The database is automatically created when instance first runs.

driver

• Ensure use correct weewx config file viz. vp2.conf.
• When prompted, select the Vantage driver.
• It also prompts for a number of vp2 station details (e.g. latitude & longitude), beyond the scope of this document.

log

debug = 3

# rsyslog weewx

# weewx vp2 collector
:programname,isequal,"weewx-vp2" /var/log/weewx/vp2.log
:programname,isequal,"weewx-vp2" stop

# weewx default

# logrotate weewx

/var/log/weewx/vp2.log \
/var/log/weewx/weewx.log
{
  create 640 weewx adm

start/stop

• Verify systemd thinks the instance has started okay.

• Leave tail running until a reporting cycle has completed, which is within 5 minutes, watching for errors.
• Verify from log details that reports have been generated without errors.

• Verify expected files have been generated.

Web-browsing to the public_html/vp2/ content is outside scope of this document.

The vp2 instance can be stopped by:

Second weewx instance: gw1000 (aka ecowitt) collector

skin

    [[gw1000Report]]          # optional for collector
        # skin for the gw1000 instance
        skin = gw1000
        enable = true         # enable gw1000 custom report
        HTML_ROOT = public_html/gw1000  # subdir

    [[SeasonsReport]]
        # The SeasonsReport uses the 'Seasons' skin, which contains the
        # images, templates and plots for the report.
        skin = Seasons
        enable = false        # disable standard report

• Update skin files for collector-specific reports – outside the scope of this document.

database

    [[archive_sqlite]]
        database_name = gw1000.sdb
        database_type = SQLite

driver

Unlike Vantage driver, which is part of the weewx distribution, GW1000 driver is an extension. We downloaded it earlier.

• Use correct weewx config file viz. gw1000.conf
• When prompted, select GW1000 driver.
• Provide station-specific details such as location.

• The sensor mappings need to be updated by editing gw1000.conf. This is outside scope of this document.

log

debug = 3

# rsyslog weewx

# weewx gw1000 collector
:programname,isequal,"weewx-gw1000" /var/log/weewx/gw1000.log
:programname,isequal,"weewx-gw1000" stop

# weewx vp2 collector

# logrotate weewx

/var/log/weewx/gw1000.log \
/var/log/weewx/vp2.log \
/var/log/weewx/weewx.log

start/stop

• Verify that systemd thinks service started properly.

• Monitor for a reporting cycle and verify no errors reported.

• Verify the expected report files have been generated.

The instance can be stopped by:

Third weewx instance: editor aggregator

This instance is primarily to produce the official, consolidated report for public consumption on website. It collates data from the collector weewx instances and perhaps date from other sources such as forecast or minor inputs too small to warrant their own collector instance.

This report draws from the databases of several weewx instances:

gw1000_binding

gw1000 instance database;

vp2_binding

vp2 instance database;

wx_binding

editor instance database – this one, the default binding wx_binding

skin

    [[editorReport]]          # raison d'etre for this aggregator
        # skin for the editor instance
        skin = editor
        enable = true         # enable editor custom report
        HTML_ROOT = public_html   # not in a subdir

    [[SeasonsReport]]
        # The SeasonsReport uses the 'Seasons' skin, which contains the
        # images, templates and plots for the report.
        skin = Seasons
        enable = false        # disable standard report

• Update skin files for editor instance-specific reports – outside the scope of this document.
• Ensure the HTML_ROOT is set to top level html directory, not a subdirectory as were the collectors for their subsidiary reports.
• Data is aggregated from the various databases by specifying their data_binding in the respective tag, where wx_binding is the default tag. So, for example, the maximum outside temperature today recorded by the vp2 collector instance, which has data_binding vp2_binding in editor.conf, is $day(data_binding='vp2_binding').outTemp.max.

Details of skin and report customisation out of scope for this document.

databases

    [[archive_sqlite]]
        database_name = editor.sdb
        database_type = SQLite
    [[gw1000_sqlite]]
        database_name = gw1000.sdb
        database_type = SQLite
    [[vp2_sqlite]]
        database_name = vp2.sdb
        database_type = SQLite

   [[wx_binding]]
        # The database must match one of the sections in [Databases].
        # This is likely to be the only option you would want to change.
        database = archive_sqlite
        # The name of the table within the database
        table_name = archive
        # The manager handles aggregation of data for historical summaries
        manager = weewx.manager.DaySummaryManager
        # The schema defines the structure of the database.
        # It is *only* used when the database is created.
        schema = schemas.wview_extended.schema
    [[gw1000_binding]]
        manager = weewx.manager.Manager
        schema = schemas.wview_extended.schema
        table_name = archive
        database = gw1000_sqlite
    [[vp2_binding]]
        manager = weewx.manager.Manager
        schema = schemas.wview_extended.schema
        table_name = archive
        database = vp2_sqlite

Driver

Ultimately, something must be injecting observations into the weewx instance to keep it turning over, though this is independent of the archive interval at the end of each of which the reports are generated. This can be either a simple subsidiary input with its own driver, or the reporter instance can be merged into one of the collector instances.

log

debug = 3

# rsyslog weewx

# weewx editor reporter
:programname,isequal,"weewx-editor" /var/log/weewx/editor.log
:programname,isequal,"weewx-editor" stop

# weewx gw1000 collector

# logrotate weewx

/var/log/weewx/editor.log \
/var/log/weewx/gw1000.log \

start/stop

• Verify that systemd thinks service started properly.

• Monitor for a reporting cycle and verify no errors reported.

• Verify the expected report files have been generated.
• Verify the files have been placed in public_html directory, not in a collector subdirectory.

The instance can be stopped by:

Finally

Operationally, the weewx collector and aggregator instances can be started and stopped independently.

Semantically, the aggregator instance depends upon sufficiently current data being available in the collector instances databases, so be wary of keeping collectors offline for long while aggregator(s) are running.