Server Version 1.23.3.4707: (Docker Plexinc/PSM Public
before the Docker update about 2 days ago - my PLEX DOCKER Server used average 150-300 MB RAM - even when 2 accounts were streaming (including HW transcode).
After the update, it shows me (in Synology-Docker-App) that PLEX is use 18GB RAM. Inside the container it show’s me about 200. Synology DSM shows me about 3% of 20GB. PlexWeb showed me 0.3 %. I restarted the container. After 2-3 Minutes it was on 18GB again. No streaming or access happend. (DLNA is off)
I cleaned up docker - deleted all Containers, Images. Downloaded public again and made a fresh start (clicking again trought the config ). Now it seems to use ram fast - but also free it again when idel. But not as before. It stays about 800-1024MB RAM - inside container about 120MB) - DSM shows me only about 600 MB in use.
What’s normal on your Plex/Docker installations in RAM usage? Can you post the numbers for me to see what is “normal” in idle and in use? In official Docker Overview and inside?
Is this normal on PLEX docker? I use other Docker APPs. they go up heavly when used - but go down instant when finished.
Does the used RAM growth when more media is added? I let this question here - but saw in the forum this one: High ram usage on Synology DS218+ - #4 by ChuckPa
I just started to implement the media. Only 4TB in - 20 more coming …
i’m not scared that RAM is used - its here for exactly that. But I would like to understand what is a average / normal RAM usage so i can check from time to time - or if I have to see whats wrong when to much is eaten / shown. At least it’s special to me, that there is shown 18gb in docker app - but internal and system wider it’s not 5% of it. SWAP is also empty and was empty by the time.
Thanks for your numbers - and maybe tipps.
PS: saw that PLEX is back on native install on DSM 7. Is it possible to install and move the configs from docker to native? Just to know
@ChuckPa i use DS 920+. Official max. ram 8gb. But there is a „supported“ version of a 16gb ram that works perfectly.
But ofc. I tested with only the built in 4gb ram also. Same result.
Here a screenshot - now it‘s the other container and not plex.
But it looked the same way. After 1-2 minutes after start. The difference - the other container goes down to 200mb after finish his workload ^^
If you have more than 8GB of RAM in the DS920 then you can expect it to FAIL.
The MAXIMUM supported RAM is 8GB
Yes, DSM will see the memory but the electronics will crash or DSM & Applications will behave incorrectly (which you already seem to be experiencing) when the maximum memory specification is exceeded. You CANNOT use more than the CPU has pins for or traces on the motherboard circuitry.
Be advised, People tried this with the J34xx CPUs and had the exact same problem.
This happens because DSM looks at the SPD chip on each RAM DIMM and adds the totals WITHOUT validating it.
At some point , you can expect:
PMS crash
DSM crash
Potential volume corruption and data loss.
A poor but equivalent analogy:
“Just because you can mount a 250 gallon fuel tank in the back seat of your automobile does not mean it is safe to do so.”
I plugged in the extra ram after I saw that it is full all the time ^^ atm. nobody have time to stream from my family, so i cant show you that it happens the same in plex (it did not before the newest plexinc/psm-docker update), but I can trigger the other application a bit. (see screenshot).
To be honest. We run 2 synologys 1x5 and 1x8 years with unsupported amout of ram (one of them runs a mailserver in vm) in company. All i was expecting when putting in to much - it wont boot or it is not allocating the addresses and at least shut down. But then we had bunch of luck nothing happend until now.
Maybe you can tell me how much ram is “normal” for idle or when in use (average).
Or does this difference have something to do with these rss and VSZ? If yes - what on the picture on the right is what? VSZ is the virtual assigned memory from all librarys while rss is only from the main library right?
Free memory will always be about 100-150 MB. Linux uses it for buffers as much as possible but keeps some free. Whether you have 4GB or 8GB or 16 GB or more, only 100-150 will be “Free”.
Your memory usage here shows:
a. 761 MB used - this is what matters
b. 2.8 GB Cache (zwischengespeichert) – totally normal. (Linux using the free RAM)
The DS918+ series, which runs with 8GB of RAM, successfully supports multiple hardware transcoding streams in 8GB of RAM.
RSS = ReSident Size – How much is currently in RAM
VSZ = Virtual SiZe – How much the process could grow to
Here you see Plex running on DSM 6,
I’ve just started scanning a new server
You see the Physical Memory used in the upper right corner
You see the total Virtual possible allocation for the container.
My development system has 1 GB of RAM and runs very well.
PMS does not need “Gobs” of memory.
As a rule of thumb, I use:
Up to 1 GB for all of PMS itself (worst case with all support processes running concurrently)
500 - 1000 MB for each stream active transcoding.
On these J-series processors, the Hardware Transcoding ASIC (UHD-600 family) will max-out at 6 simultaneous high bit rate transcodes ( about 500 Mbps total video throughput )
How does that fly against what Intel, the chip manufacture says?
Any Linux kernel will read the SPD chips and add up a total.
That’s trivial.
Try and USE all of it (write a galloping page memory test. See what happens)
(we went around and around when these ApolloLake chips first came out. Folks had assumed 16GB when it’s only 8GB and their QNAP & Syno systems started crashing until they backed down to 8GB. Machines like my Atom C2538 does support 16GB)
Oversubscribing virtual RAM is even more trivial – that’s what swap is for
Is it possible Intel published bad specs for ApolloLake? It wouldn’t be the first time but, and here’s the kicker, is it guaranteed ?
Without seeing the source code to that demonstration program (the one the scripting is based on), I will not say ye or nay.
I am a developer and want to see the allocate and lock calls. ( there are a couple different ones in Linux )
If everyone wants to do it – then go for it. It is your machine afterall.
Not to be a naysayer , if I encounter a case with PMS crashing and the system memory installed is over spec – how can I support that? It’s out of spec from the git-go. Would you want to support that? (my best analogy – cars: Put a 6.2 liter engine in a Fiat)
). Now it seems to use ram fast - but also free it again when idel. But not as before. It stays about 800-1024MB RAM - inside container about 120MB) - DSM shows me only about 600 MB in use.
