URGENT! Lightsail instance immediately crashes when reaches burstable zone without using up burst allowance.

For context, I have a lightsail ubuntu instance with 2 GB RAM and 2 vCPUs. I use it as my dev server and it doesn't have any cron running in the background when I am away or anything else that runs on schedule apart from the daily automatic snapshots. This means that if I am not using it, it's suppose to be idle.

I have installed sysstat yesterday and enabled it as well before I logged off my server. Here's a screenshot of what's happening with the CPU from 9pm last night when I am completely logged off and not doing anything on the server. Unless the sysstat is so resource heavy, I do not have a clue what's keeping it busy during this graveyard times.

As commented in other post, I suspect your CPU usage is a symptom and not the cause of your issues. Crashing immediately or becoming unresponsive is more likely caused by memory than CPU. I would recommend that you setup sysstat in the OS to see how memory is tracking before the issue occurs.

If you truly believe you need support to find the source of the issue, it would be wise to open a support case to have AWS look at your instance from the backend and help find cause.

Hi, Thanks for your help so far. Please find below the output of sar for that period and screenshots of CPU peak times. The server is just LAMP webserver.

As you can see it did not log anything on 25th Nov because the server has crashed by then.

Definitely memory issue.

Notice on the left the timestamps for logging deviate from the 10 minute cadence, indication that the system was unresponsive for a period of time. Next notice that during those same periods, you see %commit jump from 96% to 150%. You were running very close to capacity and then committed an additional 54%. During that time, it evicted items that were cached in memory to use for committed and finally you'll notice that kbavail drops to 588kb. At that time the instance becomes unresponsive till you likely rebooted it at 10:53:53.

As previously suggested, your CPU usage is a symptom of a thrashing system, not the cause. Jump to the next instance size to increase memory and I bet your problems will vanish.

Alternatively, stop some of the services that you are running to reduce required memory or add some swap space. The swap space may allow you to get by without becoming unresponsive but performance would be less than optimal. With the commit jumping 54%, I think swap will not likely meet your needs.