Ticket #276 (reopened defect)

Opened 3 years ago

Last modified 3 years ago

After full disk CPU usage is at 100%

Reported by: chemical <chem@gmx.net> Assigned to: madcat
Priority: high Milestone:
Component: hydraengine Version:
Severity: normal Keywords:
Cc:

Description

Using hydranode core 0.3 release binaries for Linux.

The core is with 10k sources at 0,05 load all the time. Last night my disk ran full and the core charged me all the time: 

[2006-Jun-01 08:03:51] Error: Error saving temp file: no space left on drive
[2006-Jun-01 08:03:51] Info: Auto-pausing file 'x' due to the abov
e error.
[2006-Jun-01 08:04:00] [Statistics] Sources: 6497 | Queued: 3780 | Up:  29.26 KB/s | Down:  66.64 KB/s
[2006-Jun-01 08:04:03] Error: Error saving temp file: no space left on drive
[2006-Jun-01 08:04:03] Info: Auto-pausing file 'y' due to the above
 error.

After this, there were corruption errors in a row: 

[2006-Jun-01 08:20:40] Warning: x: Corruption found at 223744000..233
471999 (9.28 MB)
[2006-Jun-01 08:20:43] Warning: x: Corruption found at 272384000..282
111999 (9.28 MB)

And then, 4 hours later, I freed up some space and the core resumed all downloads automatically. However, the CPU load is since that time (or before, I don't know) at 100%.

Change History

06/01/06 04:19:54 changed by chemical <chem@gmx.net>

  • status changed from new to closed.
  • resolution set to invalid.

Closing ticket, false alarm. Core was hashing long time.

06/02/06 15:33:43 changed by anonymous

  • priority changed from normal to high.
  • status changed from closed to reopened.
  • resolution deleted.

I don't have an exact error describtion because it always happens when I'm not looking but after the disk ran full (it happened again) there are hashing loops (i recognized them because the core printed Corruption messages over the same ranges all the time) which never stop until a) the file is cancelled, b) the core is stopped

After b), the core crashes and the metadata is corrupted.

There was a self-initiated oops log, but I missed to copy and paste it, sorry.

Now, after restart, the metadata is resetted and the core is hashing ALL data again.

Using r2998 (final 0.3)

06/02/06 15:34:47 changed by chemical

forgot to login.