0.8.4

• The issue of auto-determined EBS size being sometimes not an integer fixed.
• Now input files in the unicorn input json can be written in the format of s3://bucket/key instead of {'bucket_name': bucket, 'object_key': key}
• command can be written in the format of a list for aesthetic purpose (e.g. [command1, command2, command3] is equivalent to command1; command2; command3)

0.8.3

• A newly introduced issue of --usergroup not working properly with deploy_unicorn/deploy_core is now fixed.
• Now one can specify mem (in GB) and cpu instead of instance_type. The most cost-effective instance type will be auto-determined.
• Now one can set behavior_on_capacity_limit to other_instance_types, in which case tibanna will try the top 10 instance types in the order of decreasing hourly cost.
• EBS size can be specified in the format of 3x, 5.5x, etc. to make it 3 (or 5.5) times the total input size.

0.8.2

• A newly introduced issues withtibanna_4dn CLI is now fixed
• One can now directly send in a command and a container image without any CWL/WDL (language = shell).
• One can now send a local/remote(http or s3) Snakemake workflow file to awsem and run it (either the whole thing, a step or multiple steps in it). (language = snakemake)
• Output target and input file dictionary keys can now be a file name instead of an argument name (must start with file://)
  • input file dictionary keys must be /data1/input, /data1/out or either /data1/shell or /data1/snakemake (depending on the language option).
• With shell / snakemake option, one can also exec into the running docker container after sshing into the EC2 instance.
• The dependency field can be in args, config or outside both in the input json.

0.8.1

• deploy_core (and deploy_unicorn) not working in a non-venv environment fixed
• local CWL/WDL files and CWL/WDL files on S3 are supported.
• new issue with opening the browser with run_workflow fixed

0.8.0

Tibanna can now be installed via pip install tibanna! (no need to git clone)

Tibanna now has its own CLI! Instead of invoke run_workflow, one should use tibanna run_workflow.

Tibanna’s API now has its own class! Instead of from core.utils import run_workflow, one should use the following.

from tibanna.core import API
API().run_workflow(…)

The API run_workflow() can now directly take an input json file as well as an input dictionary (both through input_json parameter).

The rerun CLI now has --appname_filter option exposed

The rerun_many CLI now has --appname-filter, --shutdown-min, --ebs-size, --ebs-type, --ebs-iops, --key-name, --name options exposed. The API also now has corresponding parameters.

The stat CLI now has API and both has a new parameter n (-n) that prints out the first n lines only. The option -v (--verbose) is not replaced by -l(–long)

Pony

• tibanna_4dn is now separate from tibanna and inherits from tibanna classes and uses modules from tibanna.

• Unicorn deployment

tibanna deploy_unicorn

• Pony deployment

tibanna_4dn deploy_pony

0.8.0b1

This is a test version.

0.7.0

Now Tibanna uses Python3.6 (Python2.7 is deprecated)

---

• newly introduced issue with non-list secondary output target handling fixed

• fixed the issue with top command reporting from ec2 not working any more

• now the run_workflow function does not later the original input dictionary

• auto-terminates instance when CPU utilization is zero (inactivity) for an hour (mostly due to aws-related issue but could be others).

• The rerun function with a run name that contains a uuid at the end(to differentiate identical run names) now removes it from run_name before adding another uuid.

• Pony

  • md5/filesize for extra files nor working - fixed
  • Tibanna initiator now uses data environment
  • fdn connection exception handling is added to Tibanna initiator
  • A newly introduced error of updating wfr status upon awsem error not working is now fixed
  • spot instance capacity / instance limit handling options added

0.6.1

• Default public bucket support is depricated now, since it also allows access to all buckets in one's own account. The users must specify buckets at deployment, even for public buckets. If the user doesn't specify any bucket, the deployed Tibanna will only have access to the public tibanna test buckets of the 4dn account.
• A newly introduced issue of rerun with no run_name in config fixed.

Pony

• Newly introduced issue with bed2multivec handling at update_ffmeta_awsem now fixed

0.6.0

• The input json can now be simplified.

  • app_name, app_version, input_parameters, secondary_output_target, secondary_files fields can now be omitted (now optional)
  • instance_type, ebs_size, EBS_optimized can be omitted if benchmark is provided (app_name is a required field to use benchmark)
  • ebs_type, ebs_iops, shutdown_min can be omitted if using default ('gp2', '', 'now', respectively)
  • password and key_name can be omitted if user doesn't care to ssh into running/failed instances

• issue with rerun with a short run name containing uuid now fixed

• pony

  • bedtomultivec workflow's higlass registration support added

0.5.9

• Wrong requirement of SECRET env is removed from unicorn installation
• deploy_unicorn without specified buckets also works
• deploy_unicorn now has --usergroup option
• cloud metric statistics aggregation with runs > 24 hr now fixed
• invoke -l lists all invoke commands
• invoke add_user, invoke list and invoke users added
• log() function not assuming default step function fixed
• fixed invoke log working only for currently running jobs

pony

• app_version auto-inserted to workflow run title
• output_quality_metric field removed from workflow run objects
• better error message for update_ffmeta for AWSEM error