Farmer bloats to 32GB RAM and then stops

I was able to sync with a plot size of 4T successfully but now I can’t get the farmer to work, it just gorges itself with 32 GB of RAM and then throws the following in the logs.

2022-06-08T10:03:37.063871Z  INFO subspace_farmer::farming: Salt will be updated, recommitting in background next_salt=79f25142c329a0d0
2022-06-08T10:03:37.063982Z  INFO subspace_farmer::farming: Salt will be updated, recommitting in background next_salt=79f25142c329a0d0
2022-06-08T10:03:37.063860Z  INFO subspace_farmer::farming: Salt updated, recommitting in background new_salt=0000000000000000
2022-06-08T10:03:37.064083Z  INFO subspace_farmer::farming: Finished recommitment salt=0000000000000000 took_seconds=0.00022365500626619905
2022-06-08T10:03:37.064096Z  INFO subspace_farmer::farming: Salt will be updated, recommitting in background next_salt=79f25142c329a0d0
2022-06-08T10:03:37.063821Z  INFO subspace_farmer::farming: Salt updated, recommitting in background new_salt=0000000000000000
2022-06-08T10:03:37.064154Z  INFO subspace_farmer::farming: Finished recommitment salt=0000000000000000 took_seconds=0.0003373810031916946
2022-06-08T10:03:37.064169Z  INFO subspace_farmer::farming: Salt will be updated, recommitting in background next_salt=79f25142c329a0d0
2022-06-08T10:07:33.115613Z  INFO subspace_farmer::farming: Successfully signed reward hash 0x845398e1de9f515d093104041bdb2bf9c4422ac12d1b14e792ac5be3b8da0cba
2022-06-08T10:17:06.180733Z  INFO subspace_farmer::farming: Successfully signed reward hash 0xd887d366200e97dbfb54791bc226ee3bac98aee0eb7f2470c3e97e6739d593b6
2022-06-08T10:17:24.130810Z  INFO subspace_farmer::farming: Successfully signed reward hash 0x9ffcf707e15974a3e4ccb09f0ab93ca8bbcebd2bea0e761c8f1166ed530146d8
2022-06-08T10:17:38.088642Z  INFO subspace_farmer::farming: Successfully signed reward hash 0xc2f3c670be7b7772f5dfb614347a9f7b701de471c97c82f15e53ff27bbaeeef0
2022-06-08T10:26:24.186605Z  INFO subspace_farmer::farming: Successfully signed reward hash 0x8afdcb308c89f0a83c4df847a5c316063157c2f230aa18511ba05ac3ec3480cb
2022-06-08T10:29:26.594383Z  INFO subspace_farmer::farming: Successfully signed reward hash 0x7e922570dba13a42ab9135184cf62c256bc6a30bfd3f57d6dd39dcfc5d2b5aca
2022-06-08T11:29:06.364433Z ERROR jsonrpsee_core::client::async_client::helpers: Dropping subscription Str("JiQCCiUl0HQbdhhc") error: TrySendError { kind: Full }

This is happening every time I start the farmer. I’m using subspace-farmer-ubuntu-x86_64-gemini-1b-2022-june-05 on Ubuntu 20.04.4 LTS

Thanks for your help

Somehow it working fine in the last couple of hours. I’ll post here when this happens again.

I’ve seen the RAM usage go up like when the following happens:

  • Node completes sync and initial plot complete
  • After sync completes, the farmer begins the salt update process
  • The salt update process requires a lot of I/O throughput. If the farm is on a single HDD, it will start chugging because it’s trying to keep in sync and update salt at the same time (my guess, just an observation)
  • The farm starts to fall behind causing memory usage to increase steadily.

How were you able to sync a 4T node? How long did this take? Single HDD?

Thanks teslak. Yes I think the instances I saw where after the salt update and I do see the node getting behind during the updates.

It took me about 3 days to plot the 4T, with some idle time in the middle where syncing wasn’t progressing. It’s my old chia HDD plotting machine with 10x 10K RPM HDDs in RAID0.

Thanks! So yeah, a 10x 10K RPM RAID array would seem to be fast enough, at least eventually.

See my other post about syncing a 2TB node on an SSD - it was one of my leftover Chia plotting SSDs:

I’m really skeptical that a single cheap external USB HDD has enough IOPs and throughput to complete plot/sync/salt update.

Hopefully this issue should be mitigated with the latest release.

Here is the PR: Recommitments improvements by nazar-pc · Pull Request #575 · subspace/subspace · GitHub