Server Version#: 1.19.1.2645
Player Version#: Plex Web 4.22.3
Plex Media Server Logs_2020-04-12_00-48-50.zip (4.4 MB)
At start, I restarted PMS docker, at the time my server used around 1.8 GB ram. After watching half day live TV, the total ram increase to 2.66 GB ram. During the period I didn’t do anything, but watching live TV. I can directly see the ram slowly growing with time by monitor through htop and glances output. Each time I close the live TV, PMS only free partial ram.
No one responds?
Sorry, but 1am here on the holiday weekend.
Can this wait until Monday?
I happened to be writing a support doc.
Of course this can wait, as long as one is following this issue.
Can you think of / find a way to recreate some of what you see?
Is it confined to an activity (e.g. transcoding) or further refined (if transcoding) to a particular codec?
Anything you can help provide helps me. There is so much which could go wrong and “leak” it’s really pretty impossible to find it without knowing where to look.
What I see? Is this refers to my live TV content, or what I see on my server?
From my server, I can see it’s transcoding.
I’m about to sleep here.
How do you substantiate “memory leak”.
What evidence do you have?
I need to have some technical information to work with.
Can you help me by providing something like that?
No smoking gun here. Only point out that the memory of PMS keeps increasing with time. It’s very hard to have solid evidence confirm memory leak.
Have a good night.
Hi, what’s the status now?
The memory still continue growing. Around Monday, PMS’s memory has increased to 1 GB. As you can see the screenshot on Apr 12, the PMS’s memory was only 572M.
But unfortunately I don’t have screenshot or data now, because I accidentally unplugged the server power at that day, PMS have been restarted all over again.
Everything here is still running fine. PMS isn’t growing, memory isn’t leaking otherwise.
There is, as you stated, no smoking gun until such time as a pattern emerges.
The only thing anyone can do is keep notes and see if that pattern emerges.
Once a pattern does to limit the scope then more progress can be made.
I think I can 100% reproduce this problem. When there is time to deal it, I will give you detailed data. And, I think programmers should check their code, and there are many tools can help them to check memory leak.
BTW, I only find this problem in live TV.
Are you a coder or a developer?
Insulting is not going to gain any points here.
I don’t see why I’m insulting? I only point out some fact, is it hurt to someone?
I don’t think so, unless you think the fact is hurt.
The fact is I can always reproduce this problem, and you tell me PMS is still running fine, which is not the case in my opinion.
And I think the code need to be check, now you say it’s reviewed and checked, which is good.
My apology I can’t see insulting here, program always have bugs, it’s normal thing.
Are you mean I said the code need to be checked means programmer isn’t good enough, or something else?
I can’t see why it has that kind of meaning. Like I said, it’s a normal thing to me.
BTW I don’t need to gain point, I need problem go away.
How do you reproduce the problem ?
I have descried some above. I’ll double check all the steps and collect the memory data, and give you all the data and steps.
But you have to wait some days. Currently, I have some other things on my hand.
This looks like normal behavior, if there was a leak you’d see the memory usage continue to climb and then dip into the swap and then Plex would get OOM killed.
It’s been a week. I collected some data in these two days.
Let’s see first screenshot.
In this screenshot, you can see the uptime is 9 days. Last time I said I accidentally unplugged the power, the uptime begin with this exactly day. During those 9 days, I didn’t use plex at all, but even though the memory of PMS still grows to around 851M.
The reproduce step is easy:
Next screenshot, in this screenshot I restarted server and PMS. You can see in this period, the memory usage was very low.
Note: in this screenshot, the uptime is around 13 hours.
