GalaxyAdmins BoF

GCC2014 BoFs!

GalaxyAdmins

This page describes the GalaxyAdminsBirds of a Feather meetup being held at GCC2014.

When: Wednesday, July 2, 5:45pm

Where: Multipurpose Room 324

Contact: Dave Clements

Description

GalaxyAdmins is a group of people that are responsible for administering large Galaxy instances. We meet online every other month and at events like GCC2014, where a lot of us happen to be.

GCC2014 coincides with the two year anniversary of GalaxyAdmins starting up. The GalaxyAdmins BoF at GCC2013 was very well attended and resulted in several actions items, many of which have now been implemented. However, the past 12 months have been less successful at having this group meet online every other month.

This meetup will discuss last year's action items, what we can about meetups in the coming year, GalaxyAdmins leadership, and whatever else participants want to talk about.

Audience

If you manage, or are about to manage, a Galaxy installation, then this is an excellent group to become involved with, and this BoF is a rare opportunity to meet your fellow Galaxy managers in person.

When and Where

When Wednesday, July 2, 4:45pm
Where TBD

Attendees

If you are interested, please add your name below and/or send an email to Dave Clements

  • Dave Clements
  • Hans-Rudolf Hotz
  • Sebastian Schaaf
  • Helena Rasche
  • Dan Blankenberg
  • Christine Walls
  • Alistair Chilcott
  • Nathan Cole
  • Peter Cock
  • David van Enckevort
  • Leon Mei
  • Melissa Cline
  • Dannon Baker
  • James Taylor
  • Kyle Elrott
  • Nikolay Vazov
  • Derek Benson
  • Youri Hoogstrate
  • Ravi Alla
  • Shane Sturrock
  • Katerina Michalickova
  • Maria Doyle
  • Calogero Zarbo
  • Ravi Sanka
  • Alistair Chilcott
  • Ulf Schaefer
  • JJ Johnson
  • Ira Cooke
  • John Chilton
  • Nuria Lozano-Garcia
  • Eric Petkau
  • Carsten Rundsten
  • Dustin Cram
  • Carrie Ganote
  • Stef van Lieshout
  • Christian Rausch
  • Raj Ayyampalayam
  • Mikael Loaec
  • Brian Whalen
  • Peter van Heusden
  • Mark Rose
  • ...

There were about 40 people in attendance. We started by describing the history of the group, and then everyone briefly introduced themselves.

Discussion focused on a few different areas:

Action Items from GCC2013 Meetup

Last year's meetup provided a number of actions. Most of them were acted upon.

Separate Mailing List for GalaxyAdmins

We investigated splitting the Galaxy-Dev Mailing List into two: one for development and one for administration. We also investigated using Mailman tagging to flag GalaxyAdmins threads. Both were rejected as ineffective. People were already frequently posting to the wrong list. Adding a third list to the mix would make that problem worse and anyone subscribing to -dev or -admin would still have to subscribe to both.

Better Enable Sharing of Information

Two items were implemented late in 2013:

  • Galaxy Log Board: (since moved to the Galaxy Blog a place for linking to and/or describing experiences administering Galaxy. These entries are dated, may not be particularly maintained once created. And, more often then not, Log Board entries are just metadata, plus a link to an external resource on the web, such as a Wordpress page.
  • Galaxy Deployment Catalog (since dropped): Describes how different Galaxy Servers are implemented. This is a means for admins to learn what others have done to address particular types of demand and communities.

The GalaxyAdmins Group Itself

Online Meetups

This is the third GCC we've had in-person meetings at. In the 12 months since GCC2013, we have had several online meetups. However, due to lack of time to set up calls and line up presenters, there have been no online meetups in 2014. And, only 2 people seemed to notice this abscence.

In 2012 and 2013 we used the University of Iowa's Blackboard system to hold and record the calls. We settled on this solution because it scaled to the sizes we needed, and we could record and publish the sessions for later viewing. We used this system because the group's founder (Ann Black-Zeigelbein) worked at Iowa, and because her successor, Srinivas Maddhi, was graciously willing to continue to support this. Blackboard worked, but it was not without bumps, and it was difficult to set up.

Q: Should we restart the bi-monthly online meetups?

A. Yes.

There was enough interest in the group that the core team will again put time into making these happen. We will investigate other online meeting options for holding and, hopefully, recording the meetups.

Governance and Leadership

This was on the agenda in 2013 and was left unresolved. The Galaxy Project is keen to support the GalaxyAdmins group (for example with team member's time), but strongly believes the group should be led by members of the community.

The challenge here, as it was in 2013, is that everyone already has a day job, and putting time into running a user group may be hard to justify compared to other priorities.

Participants were again asked to consider taking a leadership role in GalaxyAdmins. In the meantime (and after we have non-core team leadership), the core team will continue with support duties.

Communication Channels

Current

The Galaxy-Dev mailing list is the primary channel for communicating with and amongst Galaxy administrators. There is also the low-volume Galaxy-Announce mailing list for big announcements.

The challenge with Galaxy-Dev is that it's a firehose: there is a lot of email and it can be a challenge to notice or find the bits that are relevant to you. It's also a mailing list, and while conversatons do happen, they are relatively slow compared to some other methods. Conversations are also presented as chronological threads, or trees at best. It can be hard to identify where the best answer is.

There is also the galaxyproject IRC channel. IRC is a much more conversational medium. However, it isn't threaded in any way, and it isn't archived and searchable anywhere. (Wait, maybe it is. See below.)

A significant percentage of attendees were not familiar with Galaxy's search engines, the Galaxy-Announce list, or the IRC channel.

Remedies

Galaxy-Dev is going to remain a firehose of information. The Galaxy Web Search and Galaxy Mailing List Search both make all Galaxy mailing lists (and the Galaxy Biostar, now help.galaxyproject.org) easily searchable. They both use Google Custom Search. They don't address the problem of finding the best answer buried in a long thread.

The Galaxy-Announce list is a good resource for keeping up with what's new in Galaxy. It only gets 3-5 emails per month.

We aren't sure if the IRC channel is currently archived and searchable anywhere. If it is not, we will take steps to archive it and make it searchable. Furthermore, we will add it to the project's search engines and include pointers to it wherever appropriate in the wiki.

Not sure what to do to make more people aware of these resources.

Common Pain Points for Galaxy Administrators

Authentication

Supporting mixed and remote local users is a challenge.

There is no support for logging out in Galaxy.

Using external services like UCSC and IGV without also leaving your server more open than you would like is difficult.

Several possible approaches were floated:

  • CI Logon
  • Support for multiple authentication methods on the same server.
  • Allow admin to specify python functions to handle login and logout.

Summary:

  • Admins want easier support for more diverse authentication methods.

Tools

Installation and Update

  • Would like to have a command line interface for ToolShed updates.
  • Would like to use ToolShed to install the tools, but would like option to do own dependency management

Tool Quality

  • Most tools in ToolShed are not labelled as verified.
  • IUC: Goal is to get them where they are supposed to be, but that's an incredible task
  • IUC decided this morning that tools that have been manually reviewed or mechanically reveiwed get a gold star. Recommendation was that the default be that only show reviewed tools are shown, unless otherwise requested.
  • There is a user rating system, but it is not widely used. Everyone in the room was encouraged to start rating repos that they have installed or tried to install.

Toolbox Filters

  • Feature is there to allow end users to say "hide this type of tool when I login to Galaxy." However, it currently does not work with external authentication.
  • Could use Tool Shed category tags to give user a list of tool domains to hide. https://trello.com/c/smf8TvXS

Admin Roles + Data Libraries

  • Make updating data libraries easier.
  • Want ability to assign specific admin privileges based on roles.
  • Authorization engine is currently pluggable, although we don't know of anyone using it.

Monitoring / Billing

  • Run time metrics are actively being worked on by core team
  • There is support for running jobs as specific users, but it doesn't work with LSF (works with other DRMAA based schedulers). https://trello.com/c/5moWcFEd
  • A common hack to track users and projects is to encode that information in the job names and then write scripts to parse that information at run time or from logs.

User Resource Allocation and Control

  • Users can't request a particular pattern of resources.
  • A basic version of this has been implemented in BWA on the test server. Users can now specify to run BWA on Stampede, which has a longer wait time, but a larger allocation.
  • However, this raises a philosophical question: Galaxy is meant to hide this level of details from the user.

Action Items

  • Everybody in the room: Consider taking on a leadership role in GalaxyAdmins
  • Everybody in the room: Start rating tools that you have experience with in the ToolShed.
  • Core Team: Create or publish location of searchable IRC channel archive; add to Galaxy searches. A searchable archive of the #galxayproject IRC channel was created in October 2014.
  • Core Team: Implement support to make authentication easier and to support more diverse, and multiple, parallel authentication methods.
  • BoF participants will contribute to or create Trello cards for issues raised during this BoF that they care about.