error: no reads in first split: bad BAM file or tiny split size?

[astewart-twist]

I keep receiving the following error while trying to convert a bam file into adam format.

java.io.IOException: 'file:/my.bam': no reads in first split: bad BAM file or tiny split size?

Additional info:

• I'm able to run the small.sam test set
• I've regenerated my bam file multiple times and checked for any kind of corruption, etc
• I've repeated this experiment on both a local environment (spark in single machine mode), and in EMR.
• spark-1.5.2-bin-hadoop2.6
• adam-distribution_2.10-0.18.2
• scala version 2.10.6

Full stack trace: https://gist.github.com/astewart-twist/ceb36e492f96fe2a2500

[heuermh]

@fnothaft @jpdna I was never able to reproduce this, might it be related to #962?

[fnothaft]

Nah, #962 would be output only, not input.

[heuermh]

Ah right, sorry, wasn't looking closely enough

[heuermh]

Fixed by HadoopGenomics/Hadoop-BAM#80 upstream, will have to wait for an update of Hadoop-BAM dependency version to 7.4.1 (or later as appropriate) to pick it up.

[heuermh]

See #989 which uses Hadoop-BAM version 7.4.1-SNAPSHOT to pick up the fix for this issue. Currently blocked on Jenkins environment changes.

[fnothaft]

@shaneknapp should be back online tomorrow. I looked and alas, I do not have permissions to make the @AmplabJenkins changes.

[heuermh]

May be fixed by #989, please retest if possible.

[fnothaft]

Closing as fixed by #989.