minimap2 gets killed, but it doesn't explain why

Question

[michael.flower.14]

I've been using bowtie2 fine aligning some reads to chr4.

I'm now trying to align to a custom reference that includes several thousand synthetic sequences. It works fine locally when I run it on my own machine, but it fails when I run it remotely on my server (aws t2.large, 8 GB RAM).

Here's the output I get:

$ minimap2 -ax map-ont --sam-hit-only --MD -Y "$REF" "$R1" > "${OUT}/${FN}.sam"
[M::mm_idx_gen::12.568*1.61] collected minimizers
[M::mm_idx_gen::28.109*1.34] sorted minimizers
[M::main::28.120*1.34] loaded/built the index for 70014 target sequence(s)
[M::mm_mapopt_update::28.120*1.34] mid_occ = 1000000
[M::mm_idx_stat] kmer size: 15; skip: 10; is_hpc: 0; #seq: 70014
[M::mm_idx_stat::28.120*1.34] distinct minimizers: 607 (1.65% are singletons); average occurrences: 353381.712; average spacing: 3.393; total length: 727835538
Killed

I've tried reading around, and also can't find a method to run it in a more verbose mode. I suspect it's hitting a ceiling in processing power? Can anyone help me work out what the issue is, and how to resolve?

[Michael]

You're running minimap2 not bowtie, but anyway, you are likely running out of memory because 8GB isn't much especially with larger reference. You need a larger instance. Also, when a process is killed by the kernel you never get an explanation, but it is almost always because of RAM issues.

[michael.flower.14]

Yes, sorry minimap2!

Answer 1

[ATpoint]

It's minimap2, not bowtie2, and you might simply run out of memory. Check memory usage locally, then you know.

[i.sudbery]

The other point to make here is that its not telling you why it has been killed because the "Killed" message is coming from the operation system, not minimap2.

Answer 2

[colindaven]

It's a RAM problem, as others have said. 8 GB is too little. I use 30-60 GB for my minimap2 processes for long reads vs plant genomes.

You can check RAM usage on your server using a variety of tools, like htop, glances, and my current fave bottom https://github.com/ClementTsang/bottom .