Someone ran into a case where the merge_alignments() job was asking for 400Gb of disk, which was too much for their cluster.
It looks like this job is using a rough estimate of size(chained(alignments)) ~= 5*(size(alignments)) which seems like it could be conservative, so this PR changes merge_alignments to exactly compute its own disk requirement, like some other jobs already do.
Someone ran into a case where the
merge_alignments()job was asking for 400Gb of disk, which was too much for their cluster.It looks like this job is using a rough estimate of
size(chained(alignments)) ~= 5*(size(alignments))which seems like it could be conservative, so this PR changesmerge_alignmentsto exactly compute its own disk requirement, like some other jobs already do.