Frequently asked questions
Short answers about TaskBeacon. For setup steps see the Docs; for common problems with scheduled jobs, the use cases.
The product
What is TaskBeacon?
A service that watches scheduled jobs. Your script sends a short web request (a ping) when it starts and when it ends. TaskBeacon knows when each run is expected and alerts you when a job fails, is missed, runs late or sends odd data, such as far fewer rows than usual. It can also check websites and read the runs of Power BI, SharePoint, dbt Cloud, Fivetran and Azure Data Factory.
Who is it for?
Analysts and small data teams whose reports and imports run on their own PC or a server, started by Windows Task Scheduler, cron or SQL Server Agent: Python, R, Stata, SAS, SPSS, PowerShell and the like. See the use cases for the problems it solves.
Why isn't Task Scheduler or cron enough?
A scheduler knows only the exit code of what it started, and only if the job ran. Stata in batch mode exits with 0 after an error, a job that loads 0 rows also exits with 0, and a run that never started leaves no code at all. See Task Scheduler shows 0x0 but the script failed.
What do I need to install?
Nothing. The script makes an HTTP call with curl, PowerShell or the language's own library.
What do Failed, Missed, Late and Odd data mean?
Failed: the script reported an error. Missed: a planned run plus its grace time passed with no ping. Late: the run started and is taking much longer than usual. Odd data: the run finished, but a number broke a rule, for example the rows dropped. See Statuses.
I have dozens of jobs. Do I add them one by one?
No. Paste a list or a CSV on Import jobs (in the menu next to + New job), or create jobs with the API.
Can it check websites and APIs too?
Yes. Our server requests the address every 1 to 60 minutes and alerts when it is down, slow, or its SSL certificate or domain is about to expire. See Website and API checks and website down alerts and SSL certificate expiry alerts.
Where can I see what changed?
On What's new: every release with its date and version, newest first (also as an Atom feed). The version is in the footer of every page.
Pricing
What does it cost?
Free to start: every account is on Free (20 jobs), and for now every Pro and Team feature is open to everyone until paid plans start. Planned for later: Free stays with 20 jobs, email and Slack alerts, row-count rules and 30 days of history; Pro at 9 € a month (100 jobs, website checks every minute, SSL and domain alerts, integrations, all alert channels and rules, the monthly report, 1 status link, 12 months of history); Team at 29 € a month (500 jobs, shared projects with any number of colleagues and roles, alerts per project, a change log, the API, unbranded status pages, 24 months of history). A free plan stays, and you get at least 60 days' notice before anything costs money. See Pricing.
Do I need a credit card to start?
No. Sign up with an email address or a Google account. Nothing is charged without your explicit agreement.
Can I get a refund?
Yes. Ask within 14 days of your first payment for Pro or Team and you get all of it back, no reason needed. See the refund policy.
What happens when I reach the job limit?
Nothing stops: your jobs keep running and alerting. You just can't add another job until you delete one. For now, write to support@taskbeacon.io if you need more.
Can I take my data with me or delete my account?
Yes, on every plan: download everything we keep about you as a file, or delete the account with everything in it, both in Settings. See Privacy.
Security and privacy
Does my data leave my server?
No. Only the ping: its kind (start, finish, fail), the time, the numbers you choose to send (such as a row count) and, on a failure, the last 20 lines of output (or the error message) your script sends. We also see the sending IP address and user agent, as every web server does. The last 100 pings per job are kept on Free (1,000 for now); see Privacy.
What if IT blocks outbound HTTPS?
The machine needs to reach one address over HTTPS. Ask IT to allow that one domain; nothing has to reach in.
Is the ping address a secret?
Treat it like one: whoever has it can send pings to that one job. It can't read anything or reach your account. Keep it out of public repositories; in GitHub Actions put it in a secret.
What about error messages and log lines?
On a failure the code from the Code tab sends the error and the last 20 lines of output, so the alert says what went wrong. We keep at most 10 KB per ping. If your output can contain passwords or personal data, untick Send output lines on the Code tab: then only the error message, return code or exit code goes.
How pings work
What exactly is a ping?
A request to your job's own address, such as https://taskbeacon.io/ping/<id>. /start at the end of the address marks the start, /fail a failure, and an exit code (/0, /1 …) reports the result. Numbers go in the query (?rows=1520) or as JSON. See What can be reported.
How does TaskBeacon know a run was missed?
From the job's schedule (daily, weekly, monthly or every N hours, like Task Scheduler) and its grace time. When a planned run plus the grace passes without a ping, the job turns Missed. See Schedules.
What if the ping itself can't get through?
The code catches the error and waits at most 10 seconds, so your job carries on as if the ping weren't there. If the finishing ping never arrives, the job shows as Missed or Late, which is the right warning.
What if TaskBeacon itself is down?
Pings sent while our server is unreachable are lost, but no job is marked Missed because of that: each job simply continues with its next run. After an outage longer than 3 hours every account gets one email about it.
Why does my cron line exit with 0 when the job failed?
Because the last command in the line is the curl that reports the failure, and it succeeds. TaskBeacon still records the failure.
What is the machine field?
The computer a job runs on, set by you on the job (the last one you used is filled in). The code doesn't send the host name. When every job of one machine is missed at once, you get one alert for the machine instead of one per job.
Languages and tools
Which languages are supported?
Python, R, Stata, SAS, SPSS, SQL Server Agent, Excel VBA, PowerShell with Windows Task Scheduler, Batch (.bat), Node.js and Linux cron. Anything that can call a web address works. SAS, SPSS, SQL Server Agent and Excel VBA are not yet tested on a real installation; tell us how it went.
Python and R: what does the code use?
Only the standard library: urllib in Python, readLines in base R. On an error it sends the message (and in Python the end of the traceback) and stops the script as before. See Python and R.
Stata: why wrap the do-file in capture noisily?
Stata in batch mode exits with 0 even when the do-file stops on an error, so the do-file has to catch its own error and report it with the r() code. See Stata do-file errors in batch mode.
SAS and SPSS?
SAS uses PROC HTTP and reports a failure when SYSCC is above 4. SPSS Statistics uses its Python 3 integration and reports an error from spss.Submit. See SAS and SPSS.
PowerShell, Batch and Task Scheduler?
A small PowerShell or .bat wrapper runs any program and reports its exit code, with the last 20 lines of output when it failed. See PowerShell (Task Scheduler).
SQL Server Agent and Excel VBA?
T-SQL can't call a web address, so two extra CmdExec job steps send the pings. An Excel macro uses MSXML2.ServerXMLHTTP and closes Excel after the run. See SQL Server Agent and Excel VBA.
Linux cron and Node.js?
For cron, one crontab line around your command. For Node.js 18 or newer, the built-in fetch, no packages. See Linux cron and Node.js.
Alerts
How am I told?
By email, and if you want also in Slack, Microsoft Teams, Discord, Google Chat, ntfy or through a webhook. Right away (with reminders while the job stays broken) or once a day. See Alerts.
Can colleagues get the alerts too?
Yes. Add up to 10 extra email addresses to a job; each one confirms once before it receives alerts. Or share a project with colleagues. See Organize jobs.
Will I be flooded with alerts?
No. A job that keeps switching between broken and OK sends one "keeps changing status" alert and then pauses for an hour, a whole machine going quiet sends one alert, and you can choose one email a day instead.
Can I watch a job without getting alerts for it?
Yes. In the job's Edit page untick "Send alerts for this job". It is still watched, changes status and shows in History and the reports, but sends no email, chat message or webhook. See Alerts.
What about holidays?
Add no-alert days: nothing is planned, nothing counts as Missed and no alerts go out on those days.
Can my manager see the jobs without an account?
Yes, with a read-only status link for a project (Pro and Team, open to everyone for now): job names, statuses, the last run and 90 days of history, nothing else. See Status link.
Data checks
How does the row-count check work?
Send the row count with the finishing ping. The first time it arrives, the job gets two rules: a run is Odd data when the rows drop more than 40% below the average of the recent normal runs (after 3 of them), or are below 1. You can change both rules on the job page. See 0 rows or far fewer than usual.
Can it tell me the data is stale?
Yes. Send the newest date in your data as last_date; a rule such as "older than 1 day" turns the run Odd data. See data freshness check.
Can I check my own numbers or the output text?
Yes. Send any number (a total, a count of errors) and set a rule: below X, above X, or drops more than X%. Output rules turn a run Failed when its output contains words such as ERROR. See What can be reported.
Does TaskBeacon look at my data to do this?
No. It only sees the numbers your script sends. Counting rows or finding the newest date happens in your script, on your machine.
Another question? Write to support@taskbeacon.io.