Error: 28 (No space left on device)

[Former user]

Hi all,

I was running kma ok until suddendly I started getting the following error messege:

Error: 28 (No space left on device)

Detailed output:

Collecting k-mer scores.

Reading inputfile: ./samples/AFA016_1.fastq ./samples/AFA016_2.fastq

Phred scale: 33

Total time used for DB loading: 184.03 s.

Finding k-mer ankers

Error: 28 (No space left on device)

I do actually have physical disk space available in my working space so I guess the software is trying to write something somewhere else in the server (rhel) but I don´t know wherte I've tried changing TMPDIR to somewhere where I know for sure there is enough free space, but to no avail.

If this has to do with temporary files can we specify where these should go?

Any ideas would be much appreciated Thanks a lot!

[ptlcc]

It seems to be the same as issue #21, setting -tmp should do the trick.

Best,
Philip

[santos alonso]

OK, thank!. Seems to work now. Thanks for your helpful and prompt reply!

(I also should have read more carefully the manual!)

[santos alonso]

Sorry to insist on this issue, but although specifying the -tmp flag solved my problem before, now I am runnig out of disk space, and it seems that 4TB of free disk space is not enough to overcome the error28 issue. My question is twofold,

-is there any way to forsee how much disk space will the temporary files use?

-is there any way to minimize the size of the temporary files?

‌

Thanks!

[ptlcc]

Hi Santos

It is a bit tricky to estimate the size of tmp-files, but it should be possible to questimate it based on prior experience. The tmp-files contain all mapped reads together with the possible template matches. This means they are dependent on the number of mapping reads and the redundancy in the database. I.e. if there are a lot of closely related sequences in the DB, each read will likely have several matches.

There is not much to do about the problem on the userside, as reducing the aforementioned factors will affect computational requirements and accuracy. But reducing the redundancy in the database, as that will reduce the information needed pr. read. Splitting the database up into non-overlapping parts will also reduce the size, as fewer reads will map, this does however require mapping to several databases for each sample.

I will keep in it mind to check for optimisations on the kma-implementation side.

Best,
Philip

‌

[santos alonso]

OK, thanks a lot!