Check that the full node doesn't skip any signage points. If it does,
send a normal priority notification.

The 30 second threshold is violated multiple times per day in the
current network. It's not exactly clear to me why but further analysis
of the logs shows that no signage points are skipped which indicates
that the farmer can still participate in all challenges.

Chiadog is now monitoring signage points so we can safely increase
the threshold here and rely more on other checks.

Logs indicating what checks and services are active to help provide
confidence that everything is running.

Also add warning log in case no notification service was enabled in the
config and info log for every keep-alive check.

There's a race condition and it can happen that the log consumer tries
to call the log handler before it is fully initialized.

This will enable monitoring remote harvester(s). Pre-requisite is having
setup key-based SSH authentication with your harvester machine.

This will help to distinguish notifications when running multiple
instances of chiadog to monitor more than 1 remote harvester.

Just create multiple configs for each harvester with unique prefix and
correct IP configuration and execute:

python3 main.py --config config-1.yaml
python3 main.py --config config-2.yaml
python3 main.py --config config-3.yaml

* Downgrade keep-alive logs to DEBUG
* Make notification texts more consistent with <Problem> <Reason> format

This is scenario observed on actual network. Seems unrelated to local
node because it's observable from multiple nodes at the same time.

Add some handling to ignore these type of events and reduce the
resulting false alarms.

Might be a good idea to have these visible at least as INFO logs to
understand how the network behaves over time.

Export SHOWCASE_NOTIFICATIONS=1 on top of API token and user key and run
tests to generate presentable notifications:

python3 -m unittest
tests.notifier.test_pushover_notifier.TestPushoverNotifier.testShowcaseGoodNotifications

python3 -m unittest
tests.notifier.test_pushover_notifier.TestPushoverNotifier.testShowcaseBadNotifications

Bumps [pyyaml](https://github.com/yaml/pyyaml) from 5.3.1 to 5.4.
- [Release notes](https://github.com/yaml/pyyaml/releases)
- [Changelog](https://github.com/yaml/pyyaml/blob/master/CHANGES)
- [Commits](yaml/pyyaml@5.3.1...5.4)

Signed-off-by: dependabot[bot] <[email protected]>

Implements notifications through the Telegram API:
https://core.telegram.org/bots/api

Closes martomi#11

Run notifier integration tests only if environment variables are
provided. This makes it easier for anyone to run the full-suite of tests
without needing to register for tokens for all integrations.

There was confusion that chiadog is somehow reliant on Telegram API.

Simple Discord notifier using [webhooks](https://support.discord.com/hc/en-us/articles/228383668-Intro-to-Webhooks)

* Trying to re-use the SMTP server connection if it is still available.
* Tested using AWS SES, but any SMTP relay should work.
* Include SMTP details in README

Co-authored-by: Martin Mihaylov <[email protected]>

…i#25)

Would be useful to now if any of the notifiers are taking too long to
execute so we can potentially address the issues in the future.

Also reducing number of events per category to 1 (except for low
priority). Otherwise tests are sometimes failing due to making too many
consequetive requests to some services.

If the internet connection is interrupted and a notifier attempts to
send a notification to the user, it'll fail with uncaught exception
which prevents the watchdog process from further attempts.

This applies a critical fix that will prevent that.

Also add explicit timeouts for all network requests and switch to a less
confusing return value. True = Success instead of True == Failure.

Provides a second layer of redundancy for the use-cases when chiadog is
unable to notify you.

Declutter the main README page by providing shorter summary overview of
available integrations and move detailed instructions outside.

Closes martomi#29

To reduce noise from chiadog, individual skips that are very infrequent
and happen only a few times per day will no longer be reported through
immediate notifications. All skips will be included in the daily summary
however, and they'll also be still visible in the logs.

If there are more than 1 single skip within 60 minutes, that will be
considered an issue and reported via notification. Same for larger gaps
in signage points. The exact gap will also be reported now.

This will replace the notifications for found proof because finding a
proof is not a guarantee for a farmer reward. The found proofs will
still be tracked and reported in the daily summary.

This also has the side-effect that it'll probably also work with
transactions and not only farmer rewards.

* New notification type DAILY_STATS
* StatsManager that collects stats and sends out daily
* Configurable time of day when the summary is sent out
* Follow similar pattern to condition checkers for stat accumulators

This is now necessary because daily summary notifications are longer and
don't fit in the email subject.

* It's now possible to enable / disable daily summary notification
  individually for each integration. E.g. only receive daily summaries
  over E-mail and use Pushover only for alarms.
* Same for wallet notifications which are only informatve

For consistency and also fixes a bug that prevents * markdown style to
work because of empty space following after that without icon.

This log has been removed from chia in the most recent versions so it
serves no purpose to have the logic here. I've submitted a PR to re-add
the log back, if that happens we can revert this change.

Chia-Network/chia-blockchain#2720

closes martomi#38

Ammend: Added notes about dropped wallet feature:
martomi#40

* Add Windows support for file read and interrupt.
* Add install/start scripts for Windows and Linux

Not yet ready for release. Needs more testing and better instructions.

Fixes martomi#46
Fixes martomi#49

* Remote host discovery through uname

Made the following changes to fix a Pathlib bug where the log file could not be found in the event of a Windows-run chiadog connecting to a remote Linux harvester:

- Renamed `_get_remote_platform()` to `get_host_info()` and made it static.
- Moved `get_host_info()` from class initializer to `create_log_consumer_from_config()`
- Replaced `Path` with either `PosixPurePath` or `WindowsPurePath` depending on  remote host OS info
- Added line to `.gitignore` to allow multi-remote testing
- Updated `util.py` (code formatter output)

Expanded path needed to be passed to Popen as a string.

Closes martomi#80

Added Python check to verify python version is 3.7 or above before proceeding with installation

This reverts commit 3a6cbe2.

As of Chia 1.1.3, the log for added coins is active again so we can
restore that feature.

This is required for Windows to correctly read the files.

Add overall total number of searches. Also add percentage of total searches for 5s/15 measurements

Added support for an optional MQTT notifier.
The notifier supports optional username/password authentication
Added MQTT instructions to INTEGRATIONS.md.

Custom SSH port for remote logs can now be specified in config.yaml (issue martomi#89)

* Dowgrade log output about out-of-order events to debug log level to
  reduce questions and concerns about this (expected) phenomenon
* Add clarification to the comment explaining when it happens
* Add validity flag in the return that allows the users of the function
  to clear their state from wrong previous SP records. Thus preventing
  incorrect signage point skip notifications.

There are still unresolved known issues on Windows so add a note about
that to reduce repeated reports and questions.

Closes martomi#112

amount in debug.log is no longer quoted

* Run black, flake8 and mypy tests on creating a pull request.
* Use typing.Dict instead of dict to resolve mypy issue

…artomi#133)

Without this change, logs like this one were not correctly parsed:

harvester chia.harvester.harvester: INFO     10 plots were eligible for farming c599f64427... Found 0 proofs. Time: 0.45573 s. Total X plots

* Add test case for network fork
* Fix logic to handle this case properly

Fixes: martomi#116

Should result in less false alarms.

* Add `pygtail` to enable better log rotation handling. It writes an artefact file (`debug.log.offset`) to the `chiadog` folder.
* Add `retry` to catch Pygtail errors and retry
* Add `has_rotated` method which reads log filesize over SSH using Powershell
* Removed `paranoid` option from `Pygtail` - might not be required and requires considerable writes to disk. Should we still see 'harvester offline' false positives, this might be good to re-enable.
* Remove OS-specific implementations of Local FileLogConsumer

Closes martomi#147
Closes martomi#58

I see little value in making refactors and adding additional
configuration parameters for this. A threshold of 2 should resolve all
issues that are experienced during copying of plots.

Note that 2 is an instantenous threshold since the max value gets reset
after that. Meaning that plot count may drop by 10 without a
notification as long as its dropping in steps of 1.

For transparency of what is happening, I've added an info log that is
always triggered, regardless of the threshold.

The test case is adjusted to contain a scenario with decreasing plot
count by 1 which is expected not to trigger a notification.

When running an unsuccessful `black` format check the script automatically runs the formatter and commits the altered source code.

This currently results in an error:
```
Run black src tests
reformatted src/chia_log/log_consumer.py
All done! ✨ 🍰 ✨
1 file reformatted, 62 files left unchanged.
error: pathspec 'feature/improve-harvester-offline' did not match any file(s) known to git
Error: Process completed with exit code 1.
```

Moved the black format check to the back of the queue, as it can be automatically fixed, while the others require manual intervention.

Also removed the `black --check` since we want to make the changes anyhow

Currently the loop isn't throttled which creates busy-wait and maxes out
one of the CPU cores to 100%. This is because Pygtail exits its own loop
every time it reaches the end of the debug.log file.

This also makes WARNING logs about non-existing debug.log file during
log rotation statistically improbable.

Closes martomi#166