18 Feb Status Update
by Máirín Duffy
Hi everyone!
We had our weekly upstream ChRIS status meeting this morning - it is
Thursdays @ 0900 EST (1400 UTC) on Google Meet (let me know if you'd
like to be added to the invite!) You are always welcome to join this
public, upstream meeting, it is a great way to check in with other
contributors, let us know how things are going if you have anything to
report, and also a great way to get any questions you might have answered.
~m
ChRIS Status 18 Feb 2021
=====================
:Date: 18 Feb 2021
Present
-----------
- CJ Yoruk
- Rudolph Pienaar
- Gideon Pinto
- Máirín Duffy
Action Items
-----------------
- CJ will move ChRIS on MOC wiki docs over to ChRIS docs repo in asciidoc
- CJ continuing investigation on openshift access issue
- Gideon continuing to work on new panel layout for feeds in UI with
likely demo next week
- Rudolph investigating COVID-Net PDF report breakage on x86 to see if
caused by powerpc
- Mo continuing to look at team-19 revival internally at Red Hat
Status
---------
CJ
~~
- last week fixed per system volume problem - when we run the plugins,
we req openshift vol. back then was 5 gb. now only 2, 20 and 50
available. none of the 20s were available. after the fix, all of the per
system volumes we request are getting boudned. not sure if it's
meaningful to expose it as a pman parameter bc its an openshift
parameter. can specify in the json file, not sure if valuable to create
it as a parameter for pman. it's not in the new pman right now, does it
make sense to bring it back in the new pman or leave in the json.
- pman repear was causing errors in MOC, asked Parul about it. She said
we dont use it anymore and we can remove it if we want to. I tried to
remove it in the deploymnet template we have to see if its working or
not. it didn't cause any problems to take it out.
- recently started having problems accessing openshift, starting
tuesday. yesterday he couldn't reach the console itself... reached Rob
Baron and waiting on him to get back to him. Asking Sandip if he has any
problem running anything on Openshift/MOC. found some documentation that
solves the exact problem so he is going to try it - might be a cert
error. x86 cluster as of right now is not working - application has
failed or application not available.
- The documentation is at https://github.com/FNNDSC/ChRIS_on_MOC/wiki
<https://github.com/FNNDSC/ChRIS_on_MOC/wiki> - we have been using the
ChRIS docs repo....CJ willmove this over to the ChRIS docs repo in
asciidoc. If he needs write access to chris docs repo can get it from
Rudolph. There's a usecase folder, where PACS integration doc is, maybe
add it there as an MOC integration file.
Gideon
~~~~~~
- has been working on the sliding panels for the 4 up - hoping to have
something to demo next week
- has sliding, readjustable panes using PatternFly components with both
a horiz and a vert pane.
- where to put the terminal with the new layout? there is a 1:1
relationship between terminal and node, so Mo thinks it's still a good
idea to have the terminal associated with that right hand pane. Maybe if
the user scales the right hand down small enough that the terminal can't
be displayed, make it a pop out (ala currently playing youtube videos
when you navigate away.)
- One point about the sliders... if one pane is slid all the way down to
collapse, food for thought would be to pop out a button to reopen the
slider (in case user has a hard time getting it "back")
Rudolph
~~~~~~
- big update with pman - deprecating master and will move new code to
master.moving from http server in python to flaskmodule
- pman has been totally rewritten as a flask service, drastically
simplified.much simpler, more agile, busy testing now this week
- because it's a total rewrite of the old pman, the ol dpman stuff that
touched openshift no longer exists in pman.
- CJ could potentially work on this - testing it now at BCH, if it works
on their key workflows will switch later today or tomm. the change will
break the infrastructure in MOC. need to map the parts that touch swarm
to do the equivalent with openshift
- hoping this will make pman more reliable... it used to get out of sync
with jobs being scheduled by swarm or open shift. the job would run but
would lose track that it ran. as a result, at the client, the job would
seem like it never completed.
- went to COVID-Net testing meeting - wondering did pdf report break bc
of powerpc -will test on x86 to see if arch is a factor
Mo
~~
- Giving a talk at DevConf today on ChRIS / UX
(https://devconfcz2021.sched.com/event/gmM7/user-experience-ux-free-softwa...
<https://devconfcz2021.sched.com/event/gmM7/user-experience-ux-free-softwa...> )
- We are planning an internal team-19 kickoff to start things back up
again... survey went out last week and is closing tomorrow, from there
we'll probably have a better sense of current volunteer plans and how
best to kick things off, might do some internal RHT "marketing" to find
new volunteers, maybe include materials / link to the running demo so
they can play with it
--
Máirín Duffy, RHCSA
she / her / hers
Sr. Principal Interaction Designer
Red Hat <https://www.redhat.com>
300 A Street
Boston, MA USA
duffy(a)redhat.com <mailto:duffy@redhat.com>
<https://www.redhat.com>
2 years, 9 months
11 Feb Status Update
by Máirín Duffy
Hi everyone!
We had our weekly upstream ChRIS status meeting this morning - it is
Thursdays @ 0900 EST (1400 UTC) on Google Meet (let me know if you'd
like to be added to the invite!) You are always welcome to join this
public, upstream meeting, it is a great way to check in with other
contributors, let us know how things are going if you have anything to
report, and also a great way to get any questions you might have answered.
~m
ChRIS Status Feb 2021
====================
:Date: 11 Fev 2021
Present
-----------
- CJ Yoruk
- Rudolph Pienaar
- Gideon Pinto
- Máirín Duffy
Action Items
-----------------
- Mo will reach out to Red Hat licensing expert to seek advice and cc
Rudolph
- Mo will ask Audrey to add Rudolph to next Tuesday's COVID-Net user
testing meeting
Status
---------
Rudolph
~~~~~~~
- Public test server for COVID-Net UI up and running,
app.chrisproject.org <=> covidnet.chrisproject.org (same data)
- entire workflow running on IBM powerpc containers on MOC, not
openshift yet
- Jennings experimented with adding COVID-net UI embedded inside ChRIS
UI, accessible from left hand nav. Wondering about whether or not we can
do this sort of thing wrt MIT vs AGPL licensing (all of BCH ChRIS work
is MIT licensed... DarwinAI release under AGPL)... Mo will reach out to
Red Hat licensing expert to seek advice and cc Rudolph
COVID-Net UI minor issues:
^^^^^^^^^^^^^^^^^^^^^^^^^^
- Look into adding acknowledgements / contact info to footer of COVID-Net UI
- Patient MRN Patient DOB are inconsistent - doctors doing eval of the
UI will notice and hyperfocus on it, will cause issues for effectiveness
of test
- Patient says Bill but gender is female
- Patient age is 27 and then xray says 54y
- PDF report doesn't work - this is a use case in the test plan
Generally / overall we need to ensure complete consistency in the UI
before sitting medical staff down in front of it.
COVID-Net UI other issues:
^^^^^^^^^^^^^^^^^^^^^^^^^^
- UI doesn't appear to be reading the DICOM header tags into the UI,
appears to be hardcoded into the UI?
- script called tagextract, gives all the tags
- has a chris plugin that has all the covid dicom images in a container
- it appears the tags were stripped out of the headers... it is
better and accepted to replace the values with fake values instead of
completely strip out the tags
- multi user does work and would be a lot easier to set up envs for
users to test vs. having to redeploy entire CUBE app inbetween tests
- it looks like UI is hard coded to pull data in from chris user,
that is why it shows data from other folks...
- very minor code change to enable multi-user... replaced hard coded
logins with current user login
- delete feed function exists probably, but not tested / reliable. so
if we had to clean out and re provision the env before each test user,
we'd have to tear down CUBE and redeploy
- Rudolph will join next COVID-Net user testing meeting on Tuesday to
discuss
CJ
~~
- deploying and testing pfioh and pman for the past few days
- last semester had version on MOC that didnt support SSO... last
semester when they ran pfioh and pman, they were using sandip's version
because the fnndsc didn'ts upport SSO (Sandip retrofitted some stuff so
that pman and pfiohswiftstorage in openstack... Sandip has a for that
drifted from the reference
- fnndsc now supports SSO, was updated in Dec, so he is now using that
image instead of Sandips. He had to change a little bit... the
credentials, but it works fine now. Tested pushing and pulling files
frmo OpenStack
- One issue with pman, remember from last semseter, running the plugins
for each pod we were requesting per system logins for 5 GB.... talkedf
to MOC, didnt have 5GB claims anymore, switcehd to 2 GBs, testing it now
that doesnt seem to be available. Created another MOC ticket on this,
going over some docs on OpenShift/Kubernetes to understand the problem
about per system login claims....
- Good to know... pman makes the request, when we run the plugins, we
create another pod in OpenShift. For that pod we request volume for the
container. Per system. When we request that, we were requesting 5 GB
back in Nov-Dec. When we request it now it says no per system volume
available anymore. After that tried a version that requests 2 GB now.
- Where is the request done.... in the pman repo? openshiftmanager.py
of pman... YAML file
- It worked with 2 GB and then it stopped working...
- Not great that the value for the storage is hard coded.... would make
more sense to expose that up so that you could run pman with a flag to
change the value.... think about what OpenShift values we'd want to
control via CLI flags and convert pman to support that
Mo
~~
User testing for COVID-Net
^^^^^^^^^^^^^^^^^^^^^^^^^
- Concerned about provisioning accounts for user testing on
app.chrisproject.org, wondering can I use a script to provision user
test per user from my own local machine up to app.chrisproject.org
Rudolph answer: Yes, just use the command line flag to speak
app.chrisproject.org with the chris user credentials
- Wondering if the metadata in a DICOM can be modified, rpienaar has
commandline scripts to modify scripts to modify dicom metadata, can ping
him offline to obtain
- Not sure how COVID-Net team uploaded data to swift storage. Could use
UI to upload to swift storage.
PACS Query
^^^^^^^^^^^^
- When Rudolph has more cycles, let's talk about revisions to the UI
design so Mo can update the click-thru prototype. Not urgent can wait a
couple weeks if need be
Tweaks to Feed UI
^^^^^^^^^^^^^^^^^^
- Threw together some ideas for refining the ChRIS feed UI as we've been
discussing the past couple of meetings, here's one WIP:
https://raw.githubusercontent.com/FNNDSC/cube-design/master/CHRIS-UI/4%20...
<https://raw.githubusercontent.com/FNNDSC/cube-design/master/CHRIS-UI/4%20...>
- Patternfly has a new resizeable dock component we might be able to use
to make this possible.
--
Máirín Duffy, RHCSA
she / her / hers
Sr. Principal Interaction Designer
Red Hat <https://www.redhat.com>
300 A Street
Boston, MA USA
duffy(a)redhat.com <mailto:duffy@redhat.com>
<https://www.redhat.com>
2 years, 9 months
