This might be a bit obtuse as I come from a DevOps background not an IT / Infrastructure background but I absolutely love Azure Application Insights. I’m not super familiar with how much it costs tho as I’ve never used it off a company card haha. It might not make sense for server monitoring as it’s built for watching applications.
But really monitor meaningfully. CPU usage matters but a high CPU usage doesn’t indicate an issue. High load doesn’t mean an issue.
High CPU for a long period of time or outside normal time frames does mean something. High load outside normal usage times could indicate an issue. Or when the service isn’t running. Understand your key metrics and what they mean to failures, end user experience, and business expectation.
Start all projects with monitoring in mind, the earlier to you begin monitoring the easier it is to implement. Re configuring code and infrastructure after the fact is a lot of technical debt. If you are willing and can guarantee that debt will be handled at a later time then good luck. But we know how projects go.
Assign flags to calls. If your application runs results in a response that’s started from and ends up at an end user, Send an identifying flag. Let that flag travel the entire call and you are able to break down traces and find failures… Failures don’t have to be in error outs, time outs. A call that takes 10x longer than the rest of the calls can cascade and shows the inefficiency and realiability.
Spend time on log and error handling. These are your gatekeepers to troubleshooting. The more time spent upfront making them valuable, the less time you have to look at them when shit hits the fan.
Alerts and Monitors MUST mean something. Alert fatigue is real, you experience it everyday I’m sure. That email that comes in that has some kind of daily/weekly status information that gets right clicked and marked as read. That’s alert fatigue. Alerts should be made in a way that scales.
Take a Look as a time allows - logs with potential issues
Investigate as something could be wrong - warnings
Shits down fix it - Alert
APM matters
Collect that data, you want to see everything from processor to response times, latency, and performance. These metrics will help you identify not only alerting opportunities but also efficiency opportunities. We know users can be fickle. How long are people willing to sit and wait for a webpage to load…. Unlike the 1990’s 10-30 seconds is not groovy. Use the metrics and try to compare and marry them with business key performance indicators(KPI). What is the business side looking for to show things are successful. How can you use application metrics and server metrics to match their KPIs.
Custom scripts are great. They are part of the cycle that companies go through.
Custom scripts to monitor —>
Too much not enough staff —>
SAAS Solutions (Datadog, Solar Winds, Prometheus, Grafana, New Relic) —>.
Company huge SAAS costs high and doesn’t accurately monitor our own custom applications —>
and we’re back to custom scripts. Netflix, Google, Twitter all have custom monitoring tools.
Many of the SAAS solutions are low cost and have options and even free tiers. The open source solutions also have excellent and industry level tools. All solutions require the team to actively work on them in a collaborative way. Buy in is required for successful monitoring, alerting, and incident response.
My comment as someone who used to do this professionally:
4 golden signals:
Latency (how long the transaction / query / function takes)
Traffic (queries per second)
Errors (errors per second)
Saturation / Fullness (how close to maximum traffic are you. This can be I/O, how close are you to maximum bandwidth, memory: how close are you to running out of RAM, threads: how many serving threads are you using out of the total thread pool, CPU: how much free CPU capacity do you have)
Don’t veer too far from measuring those key things. You might be able to get many other rates and values, but often they’re derived from the key signals, and you’d be better off monitoring one of the golden signals instead.
3 types of alerts:
Pages, as in “Paging Doctor Frankenstein”. High priority that should interrupt someone and get them to check it out immediately.
Tickets / Bugs. These should be filed in some kind of bug reporting / ticketing system automatically. That system should be one where someone can look up the day’s or the week’s bugs and see what needs investigation, track it, and then ultimately resolve it. This level of alert is for things that are serious enough that someone should be periodically checking the bug / ticket system to see if anything needs investigation, but not important enough that someone should drop what they’re doing and look right away.
Logs. Write info to storage somewhere and keep it around in case it’s useful for someone when they’re debugging. But, don’t page anyone or create a ticket, just keep the info in case someone looks later. Graphs are basically a form of logs, just visual.
Tempting as it may be, never email. Emails just get ignored. If it’s high priority you should page. If it’s not that high priority, file a bug / ticket.
For latency, use distributions: 50th percentile latency, 90th percentile latency, 99th percentile latency, etc. Meaning for 50th percentile, half the users have this much latency or better. For 99th percentile 99% of users have this much latency or better, 1% of users have this much latency or worse. The reason for this is that an average latency is not very useful. What matters are the outliers. If 99% of operations complete in 500 ms but 1% take 50s, the average latency will still be approx 500 ms, but that one operation that takes nearly a minute can be a sign of something, either breakage or abuse.
Black and white box monitoring are both important.
White box monitoring is monitoring as someone who knows the internals of the system. Say monitoring the latency of the GetFriendsGraph() call. As someone who knows the code you know that that’s key to performance, it has a DB as a backend, but there’s a memory cache, and so-on.
Black box monitoring is monitoring viewing the system as a black box whose internals you pretend you don’t understand… So, instead of monitoring GetFriendsGraph() you monitor how long it takes to respond to loading http://friends.example.org/list/get_buddies.jsp or whatever. That will include time doing the DNS lookup, time going through the load balancer, querying the frontend, querying the backend(s) backends, and so on. When this kind of monitor experiences errors you don’t know what the cause is. It could be broken DNS, it could be broken load balancers, it could be a DB crash. What it does tell you is that this is a user-visible error. With white-box monitoring you may know that the latency on a certain call is through the roof, but the black box monitor can say that it isn’t an issue that is actually affecting most users.
In terms of graphing (say GraphViz or whatever), start by graphing the 4 golden signals for whatever seems to be important. Then, treat the graphs like logs. Don’t stare at them all the time. Refer back to them when something higher priority (bugs / tickets or alerts) indicates that something needs an investigation. If additional graphs would have helped in the investigation, add more graphs. But, don’t do it just to have things to look at. Too many graphs just becomes visual noise.
This is pretty accurate to what I do professionally.
The point made here about the Average user experience is super super important. It’s good to know what that is for several reasons. Mainly performance tuning. But when it comes to trying to prevent disasters the middle isn’t useful.
Another thing to add. This came to me recently. There are two kinds of graphs and dashboards, those for technical folks and those for managers and non-technical folks. You want to develop both or one with variables to then simplify the graphs/dashboard. Annotations and good titles IMHO are good. Some folks prefer to have technical graph titles. I get the draw but I have to deal with multiple leads, C levels, project managers, and managers that don’t care about the technical stat just where it is compared to where it should be
I haven’t looked really. There is a DevOps community I think. Haven’t seen any SRE (site reliability engineer) or monitoring communities, One will probably pop up sooner than later.
Hey all,
I just posted an update here https://sh.itjust.works/post/49787
I probably should have listened to my gut and not have went to bed last night.
Should be fixed now, it was related to an nginx setting that needed to be tuned.
I’d hope sleep takes priority.
You got some monitoring in place? Can offer some assistance with monitoring ideas if you need, is part of what I do.
Also take care of yourself. We can go outside if we can’t log in. Or go back to work…
deleted by creator
This might be a bit obtuse as I come from a DevOps background not an IT / Infrastructure background but I absolutely love Azure Application Insights. I’m not super familiar with how much it costs tho as I’ve never used it off a company card haha. It might not make sense for server monitoring as it’s built for watching applications.
I can give a brief(ish) overview sure.
Monitor everything :P
But really monitor meaningfully. CPU usage matters but a high CPU usage doesn’t indicate an issue. High load doesn’t mean an issue.
High CPU for a long period of time or outside normal time frames does mean something. High load outside normal usage times could indicate an issue. Or when the service isn’t running. Understand your key metrics and what they mean to failures, end user experience, and business expectation.
Start all projects with monitoring in mind, the earlier to you begin monitoring the easier it is to implement. Re configuring code and infrastructure after the fact is a lot of technical debt. If you are willing and can guarantee that debt will be handled at a later time then good luck. But we know how projects go.
Assign flags to calls. If your application runs results in a response that’s started from and ends up at an end user, Send an identifying flag. Let that flag travel the entire call and you are able to break down traces and find failures… Failures don’t have to be in error outs, time outs. A call that takes 10x longer than the rest of the calls can cascade and shows the inefficiency and realiability.
Spend time on log and error handling. These are your gatekeepers to troubleshooting. The more time spent upfront making them valuable, the less time you have to look at them when shit hits the fan.
Alerts and Monitors MUST mean something. Alert fatigue is real, you experience it everyday I’m sure. That email that comes in that has some kind of daily/weekly status information that gets right clicked and marked as read. That’s alert fatigue. Alerts should be made in a way that scales.
APM matters Collect that data, you want to see everything from processor to response times, latency, and performance. These metrics will help you identify not only alerting opportunities but also efficiency opportunities. We know users can be fickle. How long are people willing to sit and wait for a webpage to load…. Unlike the 1990’s 10-30 seconds is not groovy. Use the metrics and try to compare and marry them with business key performance indicators(KPI). What is the business side looking for to show things are successful. How can you use application metrics and server metrics to match their KPIs.
Custom scripts are great. They are part of the cycle that companies go through.
Custom scripts to monitor —> Too much not enough staff —> SAAS Solutions (Datadog, Solar Winds, Prometheus, Grafana, New Relic) —>. Company huge SAAS costs high and doesn’t accurately monitor our own custom applications —> and we’re back to custom scripts. Netflix, Google, Twitter all have custom monitoring tools.
Many of the SAAS solutions are low cost and have options and even free tiers. The open source solutions also have excellent and industry level tools. All solutions require the team to actively work on them in a collaborative way. Buy in is required for successful monitoring, alerting, and incident response.
Log everything, parse it all, win.
My comment as someone who used to do this professionally:
4 golden signals:
Don’t veer too far from measuring those key things. You might be able to get many other rates and values, but often they’re derived from the key signals, and you’d be better off monitoring one of the golden signals instead.
3 types of alerts:
Tempting as it may be, never email. Emails just get ignored. If it’s high priority you should page. If it’s not that high priority, file a bug / ticket.
For latency, use distributions: 50th percentile latency, 90th percentile latency, 99th percentile latency, etc. Meaning for 50th percentile, half the users have this much latency or better. For 99th percentile 99% of users have this much latency or better, 1% of users have this much latency or worse. The reason for this is that an average latency is not very useful. What matters are the outliers. If 99% of operations complete in 500 ms but 1% take 50s, the average latency will still be approx 500 ms, but that one operation that takes nearly a minute can be a sign of something, either breakage or abuse.
Black and white box monitoring are both important.
White box monitoring is monitoring as someone who knows the internals of the system. Say monitoring the latency of the GetFriendsGraph() call. As someone who knows the code you know that that’s key to performance, it has a DB as a backend, but there’s a memory cache, and so-on.
Black box monitoring is monitoring viewing the system as a black box whose internals you pretend you don’t understand… So, instead of monitoring GetFriendsGraph() you monitor how long it takes to respond to loading http://friends.example.org/list/get_buddies.jsp or whatever. That will include time doing the DNS lookup, time going through the load balancer, querying the frontend, querying the backend(s) backends, and so on. When this kind of monitor experiences errors you don’t know what the cause is. It could be broken DNS, it could be broken load balancers, it could be a DB crash. What it does tell you is that this is a user-visible error. With white-box monitoring you may know that the latency on a certain call is through the roof, but the black box monitor can say that it isn’t an issue that is actually affecting most users.
In terms of graphing (say GraphViz or whatever), start by graphing the 4 golden signals for whatever seems to be important. Then, treat the graphs like logs. Don’t stare at them all the time. Refer back to them when something higher priority (bugs / tickets or alerts) indicates that something needs an investigation. If additional graphs would have helped in the investigation, add more graphs. But, don’t do it just to have things to look at. Too many graphs just becomes visual noise.
This is pretty accurate to what I do professionally.
The point made here about the Average user experience is super super important. It’s good to know what that is for several reasons. Mainly performance tuning. But when it comes to trying to prevent disasters the middle isn’t useful.
Another thing to add. This came to me recently. There are two kinds of graphs and dashboards, those for technical folks and those for managers and non-technical folks. You want to develop both or one with variables to then simplify the graphs/dashboard. Annotations and good titles IMHO are good. Some folks prefer to have technical graph titles. I get the draw but I have to deal with multiple leads, C levels, project managers, and managers that don’t care about the technical stat just where it is compared to where it should be
Thanks for the reply. Do you know of any Fediverse community for people into things like monitoring, paging, alerts and occasional sleepless nights?
NP.
I haven’t looked really. There is a DevOps community I think. Haven’t seen any SRE (site reliability engineer) or monitoring communities, One will probably pop up sooner than later.
Just happy you have a server running. No rush, get the sleep you need, and thanks for powering things (:
bed is good! and a little downtime is good for the soul. thanks for all that you do <3