Merging rma6 file error and memory leak

Bug Reports
1

I’m having an issue merging rma6 files with MEGAN6-UE (6.24.1).

When I try to merge rma6 files, I get this output

Sample mentioned in metadata is not present in document, skipping: CSMii-PHP-12c_contam-removed.rma6
Warning: Loaded metadata, ignored 1 unknown input samples.
Sample mentioned in metadata is not present in document, skipping: CSMii-PHP-12c_contam-removed.rma6
Warning: Loaded metadata, ignored 1 unknown input samples.
Info: Opened file ‘PHP-12-test.merged.megan’ with 6,226,569 reads

This doesn’t seem to affect the file’s composition, even though it seems like from the error message that one of the merge files is being ignored (each of the inputs has ~3 million reads). I get this same error when even ‘merging’ a single rma6 file.

This wouldn’t be too much a problem getting this error, but when I try to then use the ‘compare’ feature on multiple merged rma6 files, I can’t change any of the sample colours/symbols because every time I click on a sample in the ‘SamesViewer’ I get repeated error messages saying:

“Loaded metadata, ignored 1 unknown input samples.”

image
image865×2140 113 KB

Also, in case it’s relevant, when I open the comparison file with merged files, I get an error saying

Mergedfile: embedded input files not found

And for context, these files were generated with Diamond using this command, then imported as BLAST files into MEGAN with blastx.

for f in .fasta; do b=${f%.}; echo “Running Diamond on: $b”; /diamond blastx -d /blastdb/refseq_protein -q $f -o “$b”.Diamond-p.100hits.info2020-2022.fullout.blast --unal 1 --query-cover 0.85 --subject-cover 0.85 -e 0.05 -k 100 -p 40 --sensitive; done

Memory leak bug:

Also, on an unrelated bug note that’s also causing me issues, both when running locally and on a UNIX server, MEGAN ignores the maximum memory allocation set in the MEGAN.vmoptions file. In some cases, it runs until it maxes out all available memory then crashes. On the server I’m using, I have it set to use a max of 200,000 MB on a cluster that has 394,823 MB available. It’ll try to use all 360+ GB on my large files, and I’ve noticed that the reported memory usage in the MEGAN program is not consistent with how much memory is being used when track memory usage using ‘top’. I wonder if there’s a memory leak going on?

2

Sorry that the error message is cryptic: MEGAN is complaining about a mismatch in names in the metadata. It doesn’t affect the processing of the data.

I’m not sure about a memory leak, as the bound is enforced by the JRE… It could be that additional memory “outside of the JRE” is used by the SQLITE library that MEGAN uses to work with the megan db mapping file. Were you meganizing or importing alignments when the “leaks” ouccred.

3

Is it possible to have an option to stop pop-up error messages so I can still work with the file even if there’s a naming mismatch in the metadata?

I was using the ‘import from blast’ feature when it exceeded memory caps. It most commonly occurs when I’m importing 10+ blast/fasta files together into a single rma6 file (they’re all replicates). The other way I was getting around it was using the merge feature, but with all the error messages from the merge/compare feature now, it made it difficult to work around the memory issue.

4

Try deselecting Edit->Preferences->Show Notifications. That should stop notifications from popping up. Does that solve the problem?

Also, as these messages are not that important, in the next release (later today), they will no longer be shown as a “notifications” but rather will only be reported in the Message Window.

5

Sounds good, thanks!