Hacker News Reader

Comment by IronWolve | Hacker News Reader

We just consolidated all our remote offices into a global corporate account, we all had pagerduty setups for our alarming systems, solarwinds, nagios, email, etc.

I've been using pagerduty for years, and I've been getting other admins to switch from private emails to it, so no more single point of failures. They just had to be shown how simple it was to configure and use.

I won't wake up in the middle of the night to a text, but a phone call, wakes me up. I don't even answer, I refuse the call then open the pagerduty app. Love it.

View on HN

Same here. I need a phone call to pay any attention. I use UptimeRobot and their RSS feed to pipe into a Zapier Zap which uses Twilio to call me and read me the site that is down in a British voice haha. Easy to setup, works and really cost effective!

Too many weak points of failure on that setup for an alerting system.

I don't think so for my usage and I shared it because it's cost effective and would be acceptable for a lot of devs' projects. I still get UptimeRobots notifications if Zapier or Twilio hiccups. For the phone notifications, if the system fails on a given night it's not the end of the world if I don't wake up. I can fix in the morning.

But I agree for e-commerce or mission critical applications - you need redundancy and it's worth ponying up the extra money for an advanced system.

If you can fix it in the morning, why bother at all? As long as it's not an common occurrence, it should be fine, right?

Because many times it's just about being proactive and getting things done ASAP because having your client find out first is bad.

I also prefer not starting my day with a fire. I'd rather do it off-hours.

> I don't even answer

Now if only I could get it to not leave a message.

Yeah, this kind of sound a like a UX bug. Ideally the behaviour could be changed to match the way you’re using it, which is more of a wake up call than a call based notification system.

I feel like the core product is pretty easily replicated with Freeswitch and tailored with a little configuration. Eg: Ring a phone for 30 seconds (so as not to leave a message accidentally), if answered read back the issue (using text to speech), otherwise send the phone a text and <insert whatever else you want to do>.

Having set systems like this up for numerous clients, the latter bit is usually a case of waiting a few minutes for the on call person to either accept the alert, or failing that move onto whoever the fallback is for that person.

The core feature of Pagerduty here is resilience imo. As in, i expect whatever’s issue I’m having that they will not be having. I imagine if an AWS zone is down, large scale network issue, etc, any home rolled solution is likely to be impacted too.

While I get the point, two servers running Freeswitch combined with a bit of python and a regular rsyncing of the user data would give one pretty good uptime for monitoring.

Use visual voicemail, then you can clear your voicemails without actually calling it, treat them like texts.

Reply on news.ycombinator.com