Releases: 4dn-dcic/tibanna
Releases · 4dn-dcic/tibanna
1.0.5
- Fixed the newly introduced issue of needing more ebs size because of cwltool copying output instead of moving due to symlink
- the
rerun
andrerun_many
commands now can take non-integer values (e.g. "3x") for--ebs-size
- non-admin user cannot run workflows on a tibanna step function from the group that they don't belong to (Previously they could run things, but the jobs would fail if the jobs require access to a bucket that this user didn't have access to)
1.0.4
1.0.2
1.0.1
1.0.0
Release note for v1.0.0
Note : To use the functionality of version 1.0.0, the locally installed Tibanna Python package must also be upgraded and the Tibanna unicorns must be redeployed to AWS from the upgraded package.
Awsem
- Most of the stuff needed by an Awsem instance is now Dockerized, including the AWSF scripts, Cwltool, Cromwell and docker itself. The AWSF scripts are now in Python 3.8 and the base image is Ubuntu 20.04. The docker image is
4dndcic/tibanna-awsf:<latest>
(e.g.4dndcic/tibanna-awsf:1.0.0
) and is available on Docker Hub. This image is not meant for any other use outside Tibanna Awsem instances which automatically pulls this image to to its job. - New AMI based on Ubuntu 20.04 is used with fewer components than before, since most of them are mow moved to a docker image.
- The mounted EBS directory is now
/mnt/data1
, which is symlinked to/data1
(no real difference) - The workflow / user Docker image will be now be pulled and run inside the Tibanna-AWSF Docker container and it will use the data EBS to store the docker container rather than the root EBS. The users don't have to increase
root_ebs_size
for a large Docker image any more. Instead, increasingebs_size
(data EBS size) may be necessary depending on the size of the user's Docker image. - A Tibanna unicorn will partly depend on the Tibanna repo's
awsf3
directory (instead of theawsf
directory).
log
- The log file (
<job_id>.log
) format has changed (much better now!) - The Top command output is no longer a part of the
<job_id>.log
but it in a separate<job_id>.top
and<job_id>.top_latest
files, which can be retrieves by the following commands:
tibanna log -j <job_id> -t # all top command output at 1-minute interval
tibanna log -j <job_id> -T # the latest top command output
metrics plot
- The metrics plot now includes per-process CPU and memory barcharts based on the top command outputs. These plots are useful to figure out which process or step takes a lot of CPU or memory or not using enough CPU / memory, etc.
output specification
- The output target specification in the input json (job description json) can now be a dictionary.
- A zipped output can be directly unzipped into a directory in S3 by specifying 'unzip': True in this dictionary. See example below:
"output_target": {
"my_zipfile": {
"unzip": true,
"object_prefix": "some_directory_for_zip_content/"
}
}
Commands and options
- The
stat
functionality now has an option to specify job IDs. - The new
run_batch_workflows
subcommand allows running many workflows in a batch. - The new
info
command prints out the information about each job stored in dynamoDB (complementary tostat
).
AMI
- Code that builds AMI from scratch is available now, though most users don't need it, since the Tibanna AMI is publicly available.
AWS regions
- Public Tibanna AMI for the following new regions are now available:
me-south-1
,af-south-1
,ap-east-1
andeu-south-1
. (Total 20 different AWS regions)
WDL and CWL versions
- CWL draft-3 support is now dropped.
- WDL v1.0 is now supported (through Cromwell 53.1). To continue using and old WDL draft-2 workflow, specify
language
aswdl_draft2
in the input job description json. Cromwell 53.1 supports both v1.0 and draft-2, but some old components are not supported. Try switching towdl_draft2
in such a case. For backward compatibility, we use the same Cromwell 31 that we used in previous Tibanna versions, to run WDL draft-2.
0.18.3
0.18.2
0.18.1
0.18.0
-
Support for docker images on Elastic Container Registry (ECR) (in the same region as the running Tibanna/EC2)
- Requires redeployment of tibanna to enable this feature
tibanna deploy_unicorn [-b buckets] [-g usergroup] [-s suffix]
-
The problem of throwing a wrong error message when the run fails (AccessDenied to ListExecutions operation ...) is fixed.
0.17.3
The problem of resource metrics not collected for regions other than us-east-1
is now fixed.
Metrics can be collected retroactively (to recent jobs) using tibanna plot_metrics -j <job_id>
without redeployment.
For automatic metric collection for future jobs, redeployment is required as below.
tibanna deploy_core -n check_task_awsem [-g group] [-s suffix]