Memory Leak

[ngiotta]

In the later versions of Rainbowminer, there seems to be a massive memory leak. It's using over 3.2GB RAM.

[RainbowMiner]

Which version of the powershell core are you using?

I ask this, because there is a known memory leak in Powershell Core v7.3+
In case you have installed a higher version than v7.2.x, please downgrade to a v7.2.x version. Then, the memory leak should be gone.

[ngiotta]

Good to know. I have 7.4.1.5. I will try downgrading. Thanks.

[UselessGuru]

@ngiotta
Any better with 7.2.x?

[ngiotta]

@ngiotta Any better with 7.2.x?

Nope. I uninstalled PS and installed 7.2.18 and the memory leak remains. Seems to only be on my W11 boxes. I'm just not going to run Rainbow on those machines, because it's so bad, they start crashing.

[mr-manuel]

I have the same problem with any version of PowerShell. Tried with various v5.x.x and v7.x.x. versions. Tested Windows versions are Windows 10 22H2 and Windows 11 21H2.

In total I tried it on 5 different systems with the same result. The systems have all at least 8 GB of RAM and 27500 MB of reserved page file.

Rainbox Miner is on version v4.9.1.9.

After some hours the PowerShell process reaches 33 GB until it crashes. I started it at 13:35:26 and it crashed on 22:34:52, nearly 9 hours after start.

The mining process is still running and not affected by the crash, but the coins are not switched anymore.

Event ID 2004 in the Windows Evenlog:

Windows has diagnosed that the virtual memory is insufficient. The following programs occupied the most virtual memory: powershell.exe (10944) occupied 34995109888 bytes, MsMpEng.exe (3668) occupied 271937536 bytes and SearchApp.exe (7884) occupied 116056064 bytes.

Could you implement a workaround, that the PowerShell script restarts itself, if it reaches a configured memory threshold?

[RainbowMiner]

That's strange - all my Win10 systems with Powershell Core v7.2.18 run stable at around 250MByte memory consumption. Could you please upload a Debug file? Maybe there is a faulty pool module.

[RainbowMiner]

Could you implement a workaround, that the PowerShell script restarts itself, if it reaches a configured memory threshold?

That is in fact a very good idea. I will add this workaround until we find the culprit.

[mr-manuel]

Have you a link to a documentation on how to create a debug file?

Found it in another issue:

Could you please upload a Debug file (http://localhost:4000/, click "Debug file" on the left hand side)?

Can I send you the logs via mail or Discord for privacy?

[RainbowMiner]

Can I send you the logs via mail or Discord for privacy?

Sure: send it to info@rbminer.net

[UselessGuru]

That's strange - all my Win10 systems with Powershell Core v7.2.18 run stable at around 250MByte memory consumption

I am not actively using RBM, but I can confirm that all PWSH versions I have tested so far (7.2.18, 7.3.x, 7.4.1 & 4.5.0/1/2rc) have a memory leak (approx 300MB/day) when running UG-Miner which uses a lot of similar code. This is with W11 23H2 (22631.3296)

[RoyCRoderickIII]

I dont think its RBM personally, i have both a very crappy windows laptop running Win 10 pro, and my pi running ubuntu 22.04 and neither get above 6GB of ram.usage running daily. So it has to be a miner issue, seeing i dont GPU mine at all.

[UselessGuru]

So it has to be a miner issue, seeing i dont GPU mine at all.

Nope, I even see the mem leak when just running UG-Miner in 'dry run' mode (constantly querying pools & building miner objects to get the most profitable miners, but not actually running miners).

I suspect that the mem leak has something to do with custom classes used in both projects (e.g. classes device, pool & miner).
With UG-miner I occasionally (after a few hours of flawless running) I see GC messing up and breaking custom class methods (they worked in 100s of cycles, then fail suddenly, and are working again in the next cycle).

[RoyCRoderickIII]

So it has to be a miner issue, seeing i dont GPU mine at all.

Nope, I even see the mem leak when just running UG-Miner running in 'dry run' mode (constantly querying pools & building miner objects to get the most profitable miners, but not actually running miners).

Thats kinda my point, if it was RBM in general, all of us would be experiencing it, or just the window versions etc. it must be either a miner issue or the way RBM opens/closes the miner or miner version.

[UselessGuru]

it must be either a miner issue or the way RBM opens/closes the miner or miner version.

Not in my case. The mem leak is happing regardless of miners getting launched or not. It is the processes used to get the pools & miners. Not the miners getting executed.

[mr-manuel]

I can confirm that. Only the managing process crashes, the miner itself continues to mine.