Known issues¶
Batch length setting¶
The value of Batch length greatly affects performance. The defaults are reasonable, but may not be optimal in some cases. We are in the process of determining better values for various programs and use cases.
Persistent disk not properly deleted¶
As part of its normal operation ElasticBLAST starts a persistent disk and under some circumstances it is not properly deleted. To double check and delete it, please run the commands below accordingly:
gcloud compute disks list # to list disks in GCP
gcloud compute disks delete ${DISK_NAME} # to delete relevant disks
Compute cluster not properly deleted¶
ElasticBLAST allocates a compute cluster in the cloud to perform BLAST searches. Under some circumstances it is not properly deleted. To check if the cluster is still active and delete it, please run the commands below:
gcloud container clusters list # to list GKE clusters
gcloud container clusters delete ${ELB_CLUSTER_NAME} # to delete your cluster
Files left in cloud storage¶
ElasticBLAST uses cloud storage to temporally store query sequences and internal logs and metadata so that they are easily accessible during its operation. Sometimes deleting these files after the search is not successful. To double check and delete them, please run the commands below:
gsutil ls gs://${ELB_RESULTS_BUCKET}/query_batches # to list query files
gsutil -m rm gs://${ELB_RESULTS_BUCKET}/query_batches/* # to delete query files
gsutil ls gs://${ELB_RESULTS_BUCKET}/logs # to list log files
gsutil -m rm gs://${ELB_RESULTS_BUCKET}/logs/* # to delete log files
gsutil ls gs://${ELB_RESULTS_BUCKET}/metadata # list metadata files
gsutil -m rm gs://${ELB_RESULTS_BUCKET}/logs/* # to delete metadata files
A synchronous search may shut down too early¶
When doing a synchronous search (submitted with the --sync option), ElasticBLAST is continuously probing for search status to know when the search is done. When the status check times out ElasticBLAST interprets it as search failure and shuts down the cluster.
No synchronous/asynchronous option in AWS¶
This hasn’t been tested.
Too many query batches leads to failed execution¶
ElasticBLAST divides query sequences into batches and searches them in parallel. The Batch length parameter controls the size of a single batch. If the ElasticBLAST configuration leads to more than 5,000 query batches, ElasticBLAST will exit with an error message prompting you to increase Batch length parameter to decrease the number of query batches.