Hi, did something happen on 10/08 around 12:00 Pacific Time?
When looking at my machine’s memory usage via top (is this the proper way to analyze memory usage?), I’m getting about 135MB but the Fly memory dashboard says it’s consuming about 190MB. What is accounting for the ~55MB difference?
As seen in the graph, memory consumption is higher after 10/08 w/ the same app.
TIA
Probably not… /proc/meminfo is the gold standard, but it’s not easy to interpret.†
In my opinion, there is no single-number answer to the question, “how much memory is Linux using?” It depends on how fast you need it, whether you’re spinning up additional copies of a process or something completely unrelated, …
The particular graph that you posted doesn’t appear among my own dashboards, I don’t think; could you perhaps relay the exact Grafana query that is being used? (The underlying Prometheus metrics are just pulled directly from /proc/meminfo, if I understand correctly—although they’re missing SReclaimable, which personally I think is independently useful.)
(That one is currently standing at 20MB on a small machine that I checked, and a kernel upgrade could easily alter its behavior.)
†I don’t know of a better link for this one, but search down in this gigantic text file for meminfo:…
https://www.kernel.org/doc/Documentation/filesystems/proc.txt
Thanks for the reply! That link is the memory utilization link in the metrics tab “Sign in to Your Account · Fly&viewPanel=11&from=now-24h&to=now”
The kernel update a couple months back probably caused some issues on my end.
Adding swap_memory_db helped but I can’t use suspend anymore.
It looks like Fly might have done something w/ the kernel in the past day?
I have not had the time to look into it, but I have had deployments fail due to instances running out of memory. I was using the same release tag as the one previously deployed so nothing in the code has changed.
Before and after deploy the instance idles at 40-50mb more ram usage.
are you using suspend?
I’m not these are dev instances that just stay up. Little annoying as we have had to instead the scale on the instance for what appears to be no reason.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.
