Known issues on AWS

ElasticBLAST delete failures

After ElasticBLAST starts the running (i.e.: elastic-blast status indicates that there is more than 1 job running), issuing the elastic-blast delete command will take longer than usual (e.g.: greater than 15 minutes) and could eventually fail with a message similar to:

ERROR: Cloudformation stack deletion failed with errors elasticblast-${USER}-4fd9de5e8:
The following resource(s) failed to delete: [InternetGateway, InternetGatewayAttachment, NewVPC]. .

NewVPC: The vpc 'vpc-05cf58efb27ce147e' has dependencies and cannot be deleted.
    (Service: AmazonEC2; Status Code: 400; Error Code: DependencyViolation; Request ID: 0275a4e3-2776-453b-97de-9a99fff59b08; Proxy: null).

InternetGateway: The internetGateway 'igw-062b2c03bdc4dc6b6' has dependencies and cannot be deleted.
    (Service: AmazonEC2; Status Code: 400; Error Code: DependencyViolation; Request ID: 8c2bcdf5-a5b7-4352-ab59-308818229df8; Proxy: null).

InternetGatewayAttachment: Network vpc-05cf58efb27ce147e has some mapped public address(es). Please unmap those public address(es) before detaching the gateway.
    (Service: AmazonEC2; Status Code: 400; Error Code: DependencyViolation; Request ID: 3bb80118-788b-4d42-8908-c977adb4032c; Proxy: null)

In this case, the AWS resources above need more time so their dependent resources are shutdown. These particular resources do not incur in additional charges, but their existance counts towards your account’s quota. We recommend you retry the elastic-blast delete command ~24 hours later to fully shutdown the aforementioned resources.

BLAST command killed

If an ElasticBLAST search fails and elastic-blast status --verbose reports a message similar to:

The command "time blastn -query batch_002.fa -db ref_euk_rep_genomes -num_threads 16 -out batch_002-blastn-ref_euk_rep_genomes.out -evalue 0.01 -outfmt 7" returned with exit code 137
/blast/bin/blastn: line 25:    27 Killed                  blastn.REAL "$@"
Command exited with non-zero status 137

the failure is most likely caused by at least one BLAST search job that has run out of memory. BLAST searches that produce large numbers of hits may require more memory than provided by the cloud instance type selected by ElasticBLAST. It may happen, for example, when searching large genomic databases, like RefSeq Eukaryotic Representative Genome Database.

Selecting an instance type with more memory, like r5ad.16xlarge, using Machine type configuration parameter, will fix the problem.