issue about MSA server storage space

[ioj-91]

Hello. I am posting this because I encountered the following error. It seems to be an error caused by insufficient storage space. Is there a way to fix this?

MSA server enabled: https://api.colabfold.com
MSA server authentication: no credentials provided
Checking input data.
Found 0 existing processed inputs, skipping them.
Processing 1 inputs with 1 threads.
0%| | 0/1 [00:00<?, ?it/s]Generating MSA for batch_inputs_ab_only_top0001/nb_0001.yaml with 2 protein entities.
Calling MSA server for target nb_0001 with 2 sequences
MSA server URL: https://api.colabfold.com
MSA pairing strategy: greedy
No authentication provided for MSA server
Server didn't reply with json: mkdir /mnt/scratch/mmirdit/jobs/YuTOikp-5lZxanD0b_mUo8UZ-UdSAyMfta-hPw: no space left on device0 remaining: ?]

SUBMIT: 0%| | 0/300 [elapsed: 00:05 remaining: ?]
Traceback (most recent call last):
File "/lwork01/boltz/src/boltz/main.py", line 588, in process_input
compute_msa(
File "/lwork01/boltz/src/boltz/main.py", line 470, in compute_msa
paired_msas = run_mmseqs2(
^^^^^^^^^^^^
File "/lwork01/boltz/src/boltz/data/msa/mmseqs2.py", line 215, in run_mmseqs2
raise Exception(msg)
Exception: MMseqs2 API is giving errors. Please confirm your input is a valid protein sequence. If error persists, please try again an hour later.
Failed to process batch_inputs_ab_only_top0001/nb_0001.yaml. Skipping. Error: MMseqs2 API is giving errors. Please confirm your input is a valid protein sequence. If error persists, please try again an hour later..
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:05<00:00, 5.79s/it]
Using bfloat16 Automatic Mixed Precision (AMP)
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
HPU available: False, using: 0 HPUs
/opt/conda/envs/boltz/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/logger_connector/logger_connector.py:76: Starting from v1.9.0, tensorboardX has been removed as a dependency of the pytorch_lightning package, due to potential conflicts with other packages in the ML ecosystem. For this reason, logger=True will use CSVLogger as the default logger, unless the tensorboard or tensorboardX packages are found. Please pip install lightning[extra] or one of them to enable TensorBoard support by default
(boltz) user1@G81-83-0:/lwork01/AntiBMPNN$

[milot-mirdita]

There is plenty of space and inodes left, however, I learned about ext4 directory hash-collision today...

Thanks for notifying me, I'll try to fix it later today.

[milot-mirdita]

I cleared the cache, the issue should disappear. I will need to implement some long-term fix though

[adslsars]

Confirming api.colabfold.com MSA download failures, 2026-05-20 ~09:00–11:30 (JST).

The MSA search completes (COMPLETE: 100%) but the result download breaks
mid-stream, at a different byte offset each retry:

File ".../boltz/data/msa/mmseqs2.py", line 96, in download
res = requests.get(f"{host_url}/result/download/{ID}", ...)
urllib3.exceptions.ProtocolError: ('Connection broken: IncompleteRead(2828 bytes read, 5356 more expected)')
requests.exceptions.ChunkedEncodingError: (...same...)
Exception: Too many failed attempts for the MSA generation request.

It correlates with result size (controlled test, same server, minutes apart):

• 32-residue single chain → succeeds (small MSA)
• 1291-residue single chain → fails at download
• multi-chain ~2000-residue → fails at download

This is consistent with the no space left on device report in #827 — the
search runs, but large results can't be written/served, so the download
truncates. Smaller results fit and succeed. Likely the same root cause as #827
(and possibly #820's PENDING reports).

Any ETA on the storage cleanup? Same large queries ran fine ~12h earlier
(2026-05-19 evening JST).

[milot-mirdita]

Should work again. Still need to free up some more space

[adslsars]

Confirmed. Thanks!