Howto: storage node health check > discord + email alerting script

Hi there,

based on my ideas and findings from one of my previous posts, I have published my “health check script (linux + macOS)” on GitHub. :partying_face:

Please feel free to support, subscribe or comment. I’ll release updates soon, in order to make it more self-descriptive and common to use for everyday usage.

:bulb: Target group: docker and linux / macOS users.

Changelog

until v1.9.2 - feature and improvement releases

  • added a check of the LOGMIN selected part of the log, if there are time lags larger than 3 mins between a started and finished download of GET_AUDIT; the script will alert you, in case there are time lags. the selection is done by satellite, so a count per satellite will be sent by email to have some more details for further analysis. referencing related topic in the forum.
  • when the script runs between 23:50-23:59 UTC, a push msg will be sent as a daily summary, in case discord is configured
  • optimised command line output and added extra param “-q” in order to distinguish verbose “-v” outputs from “quality checks”
  • removed PUT_REPAIR from monitoring, as it should have no effect on the suspension and/or audit score(s)
  • removed an if-clause, which prevented the estimated payouts calculation is done at any time, not only, when param “-e” was provided
  • fixed if statement to enable push message EOD, which lead to not sending the push at all
  • fixed a typo in the mail templates
  • uploaded newer example screenshots
  • optimised OK case push message format
  • success rates are only shown, when scores are below 98%; can be send detailed with new param “-o”
  • fixed date issues with leading zeros

until v1.8.0 - feature and bugfix releases

  • added support for estimated payout information for UNIX + MacOS; see README for details
  • fixed an issue with awk and date selection from the external stored log file
  • added support for external stored log files on another device than the storagenode itself
  • added: fire just one error in case “other” and “severe” are equal
  • added: pending audits: run script again automatically and fire alert with the second run
  • minor optimisation in error selection for “rpc client” issues
  • fixed the external log file selection and filtering for MacOS and cleaned the code in combination with linux handling
  • added several files to be in line with GitHub community guidelines

v1.7.0-v1.7.1 - bugfix releases

  • removed the grouping of audit and repair errors; this will also properly handle the current QUIC issues of STORJ for repair traffic
  • added alerts for repair failures; meaning they will be reported separately from audit issues
  • limited the selection of repair alerts to LOGMIN instead of LOGMAX

v1.6.9-v1.6.11 - improvement releases

  • proper error handling and alerting of the QUIC issue mentioned here: storj/storj#4688
  • added logmin and logmax configration entries
  • added option “-l” (L) to specify larger or shorter log timeframes to be selected and analyzed; e.g.: ./checks.sh -l 30 -v
  • minor: optimized i/o timeout notes in verbose mode
  • minor tweaks in readme

v1.6.1-v1.6.8 - improvement releases

  • ignoring “timeout: no recent network activity” error, linked to quick library issues (see here), but added it to the “timeout” hints, so that you are aware of it
  • implemented #30 : added space information (net, without trash and gross, including trash) from storage node data and added a warning, if space is getting overused
  • minor formatting changes to discord push message
  • added 10 day delay support for new storj version, see #26
  • fixed logical issues with notifications, where boolean values were not handled properly
  • added configurable timeframes for log selection in minutes
  • small tweaks in selection of error and severe error entries
  • minor fixes for multinode usage
  • minor general improvements

v1.6.0 - feature release

  • added storj node version check
  • added support for redirected logs
  • young nodes: in case there is no repair download startet yet, no warning will be thrown

v1.5.6-1.5.7 - several improvements

  • fixed an issue with satellite’s score handling
  • added a “severe” category, e.g. alerting for docker issues or read failures
  • changed satping frequency from static 24h to standard 1h, configurable via SATPINGFREQ in .credo settings file. the initial value will be created automatically, you do not have to update your .credo file manually!
  • added #15 : success rates “info stats” pushed to discord, when flag -d was provided
  • added pending audits warning, solving #4 issue
  • fine-tuned command-line verbose output
  • several minor optimisations

v1.5.5 - feature release

  • added settings file (automatic creation), to store last ping of satellite notification, in order to just ping once a day; solving #16: limit satellite threshold warnings to once a day
  • added environment variables path option, e.g. for crontab usage
  • added settings path option, if settings file is situated somewhere else
  • extended error filter to ignore satellite service pings and emptying trash errors
  • optimized verbose output for debugging

v1.5.4 - improvement release

  • checks, if discord.sh exists and is executable
  • fixed: in case the script is called with an absolute path, discord.sh is still expected to be in the same folder, but will be correctly executed
  • shortened threshold messages a bit

v1.5.3 - improvement release

  • verified and assured macOS support
  • added jq version check to minimum 1.6 / solved #14
  • added node names in the case of multinodes to push and email alerts / solved #17
  • added success, failure, cancellation ratios into command line outputs in verbose mode

v1.5.1-v1.5.2 - improvement releases

  • added satellite audit+suspension+online score threshold checks and alerts
  • verbose option added: enable console output while execution

v1.5.0 - feature release

  • put the configuration part into a separate config file named *.credo > completely outsourced variables from the script functions
  • added disk usage for multinodes
  • added script arguments for debugging, individual config file name/path and a more self-describing help message / script usage

v1.4.0 - feature release

  • added multinode support (running on the same machine)
  • added mail + discord on/off flags to enable or disable discord pushes and mail alerts
  • general script restructures: you only need to setup your individual settings at the beginning of the script

v1.3.4 - bugfix release

  • fixed conditional statements for push string concatenation and “i/o timeout” ignore logic

v1.3.3 - improvement release

  • improved performance: just 2 docker log selections instead of 20+
  • improved mail notifications: limited error msg selection to last 1h
  • … both covering issue #11: Prevent mail alerts from spamming the mail account
  • advice to let the script run just once per hour to be in line with the error msg selection and notification
  • added echo outputs for command line usage as a feedback on what the script is working on
  • improved discord alert formatting in case repair or get/put stats are out of threshold
  • added “(io)” hint in “no errors (io)” message to let the user know, that i/o timeouts were ignored (but sent by mail in debug mode anyway)

v1.3.2 - improvement release

  • added curl, swaks, jq library check
  • added first version of a help text
  • several code optimisations

v1.3.1 - bugfix release

  • fixed an issue with the new “ignore io timeouts, if no other error occurs” feature.

v1.3.0 - added several alerts + bug fixes + optimisations:

  • fix audit log details selection from (missing) LOG variable
  • ignore i/o timeouts (satellite service pings + single satellite connects), if audit success rate is 100% and there are no other errors
  • performance: only select audit error details in case there are audit issues
  • added renice to let the script run in low performance to not block the system
  • recognizes “no download/upload activity” within the last hour
  • general optimization of the script’s styling
  • solved #7: Add alert: on low uploads or downloads
  • solved #8: Alert on a) low thresholds for success rates + b) storage node status

v1.2.0 - added audit critical + recoverable stats + log extract

v1.1.0 - added debug modes + echo outputs for command line usage

  • added mail/discord debug mode
  • created constants section to manage your individual params
  • optimised echo output for shell command line usage

v1.0.1 - v1.0.3 - several bug fixes

  • exclude ‘pieces error: filestore error: context canceled’ from the alerts
  • fixes an issue with false positive (fatal) error alerts + logs excerpt not sent by email in case of an alert
  • docker logs tail param did not work and caused the email body not to have the error messages included; removed it

about this script

this linux shell script checks, if a storj node runs into errors and alerts the operator by discord push messages as well as emails. requires at least one storj node running with docker on linux.

features

  • multinode support :earth_africa:
  • optionally discord (as quick notifications) and/or mail (with error details) alerts :inbox_tray: :bell:
  • alerts, in case: :warning:
    • audit, suspension and/or online scores are below a threshold (storj node disqualification risk)
    • audit timeouts are recognized (pending audits; discqualification risk)
    • audit time lags: download started vs. download finished is larger than 3 mins (storj node disqualification risk)
    • a threshold of repair gets/puts and downloads/uploads are reached (storj node disqualification risk)
    • there was no get/put at all in the last hour (storj node disqualification risk)
    • any other fatal error occurs, incl. issues with docker stability
    • storj node version is outdated
    • the node is offline (docker container not started)
  • reports: :newspaper:
    • disk usage
    • success rates audits, downloads, uploads, repair up-/downloads
    • estimated payouts for today and current month
    • todays upload and download statistics
  • optimized for crontab and command line usage :computer:
  • supports redirected logs to a file
  • only requires curl, jq, bc and (optionally) swaks to run :fire:

optimzed / tested for

  • debian bullseye :penguin:
  • macos monterey :apple: (jq + swaks installed with brew)

dependencies

setting up storj system health

  1. optional: setup a webhook in the desired discord text channel
  2. optional: grab your smtp email authentication data
  3. download (or clone) a copy of discord.sh *
  4. download (or clone) a copy of storj-system-health.sh and storj-system-health.credo **
  5. optional: setup discord and mail variables in storj-system-health.credo
  6. Go nuts :rocket:

* wget https://raw.githubusercontent.com/ChaoticWeg/discord.sh/master/discord.sh

** wget https://raw.githubusercontent.com/dusselmann/storj-system-health.sh/main/storj-system-health.sh && wget https://raw.githubusercontent.com/dusselmann/storj-system-health.sh/main/storj-system-health.credo

setting up variables in *.credo

you will need to modify these variables in *.credo for your specific node and smtp mail server configuration. the *.credo file must not include comments and blank lines, the following description is just for your explanation:

## discord settings
DISCORDON=true.         # enables (true) or disables (false) discord pushes
DISCORDURL=https://discord.com/api/webhooks/...
                        # your discord webhook url

## mail settings
MAILON=true             # enables (true) or disables (false) email messages
MAILFROM=""             # your "from:" mail address
MAILTO=""               # your "to:" mail address
MAILSERVER=""           # your smtp server address
MAILUSER=""             # your user name from smtp server
MAILPASS=""             # your password from smtp server

## alerting settings
SATPINGFREQ=3600        # in case satellite scores are below threshold, 
                        # value in seconds, when next alert will be sent earliest
                        
## storj node docker names and urls
NODES=storagenode       # storage node names, multiple: separated with comma, 
                        # e.g. storagenode,storagenode-a,storagenode-b
NODEURLS=localhost:14002
                        # storage node dashboard urls, multiple: separated with comma, 
                        # e.g. localhost:14002,192.168.171.5:14002

## node data mount points
MOUNTPOINTS=/mnt/node   # your storage node mount point, multiple: separated with comma
                        # e.g. /mnt/node,/mnt/node-a,/mnt/node-b
                        # enter 'source' from the docker run command here

## specify redirected logs per node
NODELOGPATHS=/          # put your relative path + log file name here,
                        # in case you've redirected your docker logs with
                        # e.g. config.yaml: 'log.output: "/app/config/node.log"'
                        #  /                       -> for non-redirected logs
                        #  /node.log               -> for single node redirect
                        #  /,/                     -> for 2 node with non-redirected logs
                        #  /node1.log,/node2.log   -> for 2 nodes with redirects
                        #  /node.log,/             -> only 1st is redirected
                        #  /mnt/hdd1/node.log      -> full path possible, too

## log selection specifica - in alignment with cronjob settings
LOGMIN=60               # latest log horizon to have a detailled view on, in minutes
                        # -> change this, if your cronjob runs more often than 60m
LOGMAX=720              # larger log horizon for overall statistics, in minutes

make sure, your script is executable by running the following command. add ‘sudo’ at the beginning, if admin privileges are required.

chmod u+x storj-system-health.sh  # or:
sudo chmod u+x storj-system-health.sh

chmod u+x discord.sh  # or:
sudo chmod u+x discord.sh

usage

you can run the script in debug mode to force a push message to your discord channel (if enabled) although no error was found - or without the debug flag to run it in silent mode via crontab (see automation chapter).

./storj-system-health.sh -d   # for a regular discord push message or:
./storj-system-health.sh      # for silent mode

optionally you can pass another path to *.credo, in case it has another name or source:

./storj-system-health.sh -c /home/pi/anothername.credo

in order to use the estimated payout information, which looks like so:

message:  [sn1] : hdd 38.62% > OK 0.25$ / 11.77$

… you should set your crontab to be run around 23:55 UTC. You need to adjust the timing, if you have a couple of nodes and/or huge log files to be analysed: the script needs to be finished before the next full hour, ideally latest 23:59:59 UTC.

it also supports a help command for further details:

./storj-system-health.sh -h

automation with crontab

to let the health check run automatically, here’s a crontab example for linux, which runs the script each hour.

15,35,55  * * * *   pi      /home/pi/storj-checks.sh -d  > /dev/null

for macos please be aware of the following specifics:

  • use crontab -e and crontab -l, although it is depricated (for now it works)
  • you do not have to use the user name, it’s to be executed with the current user
  • use full paths to your script and credo file
  • find out your standard path with echo §PATH and set it in crontab
SHELL=/bin/sh
PATH="/opt/homebrew/opt/sqlite/bin:/opt/homebrew/bin:/opt/homebrew/sbin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin"
# UNIX:
58 *    * * *   pi      cd /home/pi/scripts/ && ./checks.sh -ev
10 7,19 * * *   pi      cd /home/pi/scripts/ && ./checks.sh -ed
# MACOS
# 58    *     *  *  *  /Users/me/checks.sh -ev >> /Users/me/Desktop/checks.txt 2>&1
# 10    7,19  *  *  *  /Users/me/checks.sh -ed -c /Users/me/my.credo >> /Users/me/Desktop/checks.txt 2>&1

contributing

issues and pull requests are welcome. for major changes, please open an issue first to discuss what you would like to change.

if you want to contact me directly, feel free to do so via discord: Discord

license

GPL-3.0

5 Likes

With versions v1.3.3. + v1.3.4 the script becomes better scalable for everyday usage:

There was a huge performance improvement recently and I managed to switch of disturbing, not to say spamming email alerts, in case an error was found.

On top of that, the script now warns in case of several thresholds missed, like repair up- / downloads (risk of getting disqualified), normal data up- / downloads and if there was no upload / download activity within the last hour.

Here’s a more or less complete list of the current features implemented:

  • emails containg an excerpt of the relevant error log message
  • if the debug mode is used, disk usage of the mounted data storage disk mount point included
  • alerts in case a threshold of repair gets/puts and downloads/uploads is reached
  • alerts if there was no get/put at all during the last hour
  • alerts in case the node is offline (docker container not started)
  • optimized for crontab and command line usage (to be continued)
2 Likes

Version v1.4.0 adds:

  • multinode support
  • mail and/or discord on/off flags

“disk usage per mount point” (in case of multinodes) needs to be extended, but does not harm the main functionality of the script (alerting in case of errors/issues).

I changed the topic to a wiki, now you can update it any time.

1 Like

Versions v1.5.0 till v1.5.2 add:

  • added satellite audit+suspension+online score threshold checks and alerts
  • added disk usage for multinodes
  • put the configuration part into a separate config file named *.credo > completely outsourced variables from the script functions
  • added script arguments for debugging, verbose mode, individual config file name/path and a more self-describing help message / script usage

Full Changelog: comparing v1.4.0…v1.5.2

Version v1.5.3 changes:

  • macOS / docker support: added + tested
  • jq version check to minimum 1.6 added / solved issue #14
  • multinodes: node names added to push and email alerts / solved issue #17
  • success, failure, cancellation ratios into command line outputs in verbose mode added
1 Like

Latest changes up to version 1.5.5 :

  • added settings file (automatic creation), to store last ping of satellite notification, in order to just ping once a day; solving #16: limit satellite threshold warnings to once a day
  • added environment variables path option, e.g. for crontab usage
  • added settings path option, if settings file is situated somewhere else
  • optimized error filter to ignore “satellite service pings” and “emptying trash warnings”
  • optimized verbose output for debugging
  • checks, if discord.sh exists and is executable
  • fixed: in case the script is called with an absolute path from another folder, discord.sh is (still) expected to be in the same folder, but will be correctly executed
  • shortened threshold messages a bit, to make it better readable

Full changelog: comparing v1.5.3 with v1.5.5

Version v1.6.0:

  • added storj node version check
  • added support for redirected logs
  • young nodes: in case there is no repair download startet yet, no warning will be thrown

Is there windows version for this one?

Unfortunately no. Not sure if it works, when you run it within a Linux VM and

  1. share your dashboard to your local network
  2. export your logs to a file, where you have access from the VM

These two things should be the only settings to be made to let it work.

A lot of things have changed since v1.6.0 > latest version is v1.6.8:

  • ignoring “timeout: no recent network activity” error, linked to quick library issues (see here), but added it to the “timeout” hints, so that you are aware of it
  • implemented #30 : added space information (net, without trash and gross, including trash) from storage node data and added a warning, if space is getting overused
  • minor formatting changes to discord push message
  • added 10 day delay support for new storj version, see #26
  • fixed logical issues with notifications, where boolean values were not handled properly
  • added configurable timeframes for log selection in minutes
  • small tweaks in selection of error and severe error entries
  • minor fixes for multinode usage
  • minor general improvements

latest improvement releases v1.6.9-v1.6.11:

  • proper error handling and alerting of the QUIC issue mentioned here: storj/storj#4688 and here
  • added logmin and logmax configration entries
  • added option “-l” (L) to specify larger or shorter log timeframes to be selected and analyzed; e.g.: ./checks.sh -l 30 -v
  • minor: optimized i/o timeout notes in verbose mode
  • minor tweaks in readme

v1.7.0-v1.7.1 - bugfix releases

  • removed the grouping of audit and repair errors; this will also properly handle the current QUIC issues of storj for repair traffic
  • added alerts for repair failures; meaning they will be reported separately from audit issues
  • limited the selection of repair alerts to LOGMIN instead of LOGMAX

example output from the verbose mode:

$ ./checks.sh -l 600 -vd
2685416 (process ID) old priority 0, new priority 19
===
 *** timestamp [03.04.2022 23:12]
 *** discord debug mode on
 *** config file loaded
 *** settings: logs from the last 600 minutes will be selected
 *** settings: satellite pings will be sent: false
===
running the script for node "sn1" (/mnt/WD1001) ..
 *** node is running        : 1
 *** disk usage             : 20.79% (incl. trash: 21.49%)
 *** satellite scores url   : localhost:1234/api/sno/satellites (OK)
 *** storj node api url     : localhost: 1234/api/sno (OK)
 *** storj version current  : installed 1.50.4
 *** storj version latest   : github 1.50.4 [2022-03-18]
 *** docker log 1440m selected : #242332
 *** docker log 600m selected : #103129
 *** info count             : #98101
 *** audit error count      : #0
 *** repair failures count  : #0
 *** fatal error count      : #0
 *** severe count           : #0
 *** other error count      : #0
 *** i/o timouts count      : #0
 *** audits                 : warn: 0.00%, crit: 0.00%, s: 100%
 *** downloads              : c: 0.76%, f: 0.66%, s: 99%
 *** uploads                : c: 0.08%, f: 0.03%, s: 100%
 *** repair downloads       : c: 0.00%, f: 0.03%, s: 100%
 *** repair uploads         : c: 0.10%, f: 0.00%, s: 100%
 *** 600 m activity : up: 12890 / down: 40962 > OK
 *** i/o timouts ignored    : false
===
 message:  [sn1] : hdd 21.49% > OK 
===
 *** discord summary push sent.
 *** discord success rates push sent.

until version v1.8.0 :

  • added support for estimated payout information for UNIX + MacOS; see README for details
message:  [sn1] : hdd 38.62% > OK 0.25$ / 11.77$

# 38.62% = hdd space usage incl. trash
# 0.25$ = today's payout until now
# 11.77$ = this month's payout until now
  • fixed an issue with awk and date selection from the external stored log file
  • added support for external stored log files on another device than the storagenode itself
  • added: fire just one error in case “other” and “severe” are equal
  • added: pending audits: run script again automatically and fire alert with the second run
  • minor optimisation in error selection for “rpc client” issues
  • fixed the external log file selection and filtering for MacOS and cleaned the code in combination with linux handling
  • added several files to be in line with GitHub community guidelines
1 Like

until v1.9.2 - feature and improvement releases:

most important new features:

  • added a check of the LOGMIN selected part of the log, if there are time lags larger than 3 mins between a started and finished download of GET_AUDIT; the script will alert you, in case there are time lags (which will quickly disqualify your node, if you miss to act!). the selection is done by satellite, so a count per satellite will be sent by email to have some more details for further analysis. referencing related topic in the forum.

  • when the script runs between 23:50-23:59 UTC, a push msg will be sent as a daily summary, in case discord is configured

  • optimised command line output and added extra param “-q” in order to distinguish verbose “-v” outputs from “quality checks”

  • success rates are only shown, when scores are below 98%; can be send detailed with new param “-o”

Other things optimised:

  • removed PUT_REPAIR from monitoring, as it should have no effect on the suspension and/or audit score(s)
  • removed an if-clause, which prevented the estimated payouts calculation is done at any time, not only, when param “-e” was provided
  • fixed if statement to enable push message EOD, which lead to not sending the push at all
  • fixed a typo in the mail templates
  • uploaded newer example screenshots
  • optimised OK case push message format
  • fixed date issues with leading zeros

cc thank you @Alexey