[BUG]: Filtering for active chunk on agent aggregate is not ideal

[s3inlc]

Version Information

dev

Hashcat

No response

Description

In the AgentAPI for filtering the active chunk only the agentId and status (=RUNNING) are tested, which may not be completely ideal. There should also be a filter for completeness to avoid having chunks with some still running state but technically being completed (i.e. progress < 10000 and order by solveTime DESC).

[novasam23]

Under which circumstances is the chunk status still RUNNING while it's already complete?

I've only found this place in the code, where (if the entire hashlist is already cracked) the progress is set to 10000 and the agent is instructed to stop while the status seems to be left as RUNNING:

server/src/inc/api/APISendProgress.php

Line 528 in 885be49

$chunk = Factory::getChunkFactory()->mset($chunk, [Chunk::CHECKPOINT => $chunk->getSkip() + $chunk->getLength(), Chunk::PROGRESS => 10000, Chunk::SPEED => 0]);

Is there any other situation where the described issue would apply?

[novasam23]

@s3inlc Determining the active chunk should also consider DConfig::CHUNK_TIMEOUT for filtering, shouldn't it? So that we only consider it active if the agent is actually currently working on it.

[s3inlc]

Your example definitely is one of the cases where this would apply, I'm not sure if there is another case. But you are also right that the timeout also should be included so that chunks not received any update are not considered as active until an agent starts working on them again.
So both filters would have to be added (progress < 10000 and the chunk timeout check as well)