2019-01-03 Check-in
- Ben Leibert
- Clay Crosby (Unlicensed)
- Felimone Amone Junior
Date
Attendees / Stakeholders
@Vidya Sampath
Notes
Broken Filters (Superset Bug)
Unfortunately, the filter on the main Dashboard do PAV unexpectedly degraded over the holiday. Whereas it would previously load and initially allow for functional selections to be made, it now displays an error before even rendering.
The Appache team had suggested that we upgrade to Superset 0.28.0rc7 in order to fix it. We've since learned, though, that this may not be wise. Craig has tried running Superset's RCs and had poor luck. Jason, meanwhile, upgraded one of Ona's instances to a .26 to .27 RC, and has had trouble subsequently upgrading it ever since. It's clear that upgrading what's intended to be a production system to a Superset RC isn't a good solution.
We can perhaps setup a parallel instance of Superset for the sole purpose of documentation and training. Datasources in our current instance of Superset which don't belong to one or more dashboards could make it difficult to import/export our Superset configurations. We have 70 such datasources total, and Ben Leibert can load them onto a junk dashboard in order to make subsequent setup of a parallel RC deployment of Superset easy. Clay Crosby (Unlicensed) will ask Jason to estimate the remaining level of effort necessary to create such a parallel Superset RC instance.
Clay Crosby (Unlicensed) pointed out a workaround for getting the filters to work for the sake of documentation. Specifically, he discovered that the filters temporarily work when added anew to a dashboard. Although the approach would be painful to use during a training session and untenable for a production system, it would allow us to take screenshots the system's intended state.
From a technical perspective, it's clear that waiting for the next Superset release would entail the least amount of work. Again, Jason is looking into when it will likely occur.
Next steps:
1) Jason is looking into Superset's release cadence. There's a slim chance that they have a release planned in time for our use.
2) Ben will discuss the timing of the upcoming documentation/training efforts, along with our associated options, with the team. At the moment, Felimone Amone Junior says that Timoteo reports no concrete date associated with the training session.
Temporal Configuration
Felimone Amone Junior volunteered to highlight every temporal coefficient which the dashboard relies on. Examples of temporal coefficients (meaning coefficients which are intended to change over time) include the population and target-coverage values. In addition to having a list of these changing coefficients, it would be helpful to identify the calculations to which they belong. (Thank you very much, Felimone Amone Junior, for volunteering to help with this! Please don't hesitate to reach out to me with any questions.)
Performance
Felimone Amone Junior says that:
A) Timoteo expects 10 - 15 folks to attend the training session.
B) The dashboard should be capable of supporting 20 concurrent users.
Ben flagged CPU utilization stats he's seen which appear contradictory. Clay Crosby (Unlicensed) will run them by Jason and let us know whether he thinks our hardware can likely support 20 concurrent users.
Coverage Tab
- Updating the materialized-view would be less work that writing ad-hoc queries in Superset, so we'll go that route.
- Ona plans to update the materialized-view so as to alleviate the need for manual management of a master facility-list. There will be no explicit correlation between the facilities referenced within the dashboards' various tabs: each will have its own set of facilities which may or may not correspond to one another. (cc: Timoteo Chaluco (Unlicensed))
- Making the above change to the materialized-view affords a good opportunity for Ona to update it with functional support for temporal configuration values, which is the plan.
- Although the updated version of the view will need district-level population values, which we don't necessarily have, it won't rely on facility-level population values (which have been difficult to track down). Felimone Amone Junior said he can provide us with the district-level population metrics.
Data-ingestion Walkthrough
Both Clay and Felimone are available for a review of the data-ingestion process at 7:00 AM PST on Tuesday or Wednesday of next week. The same applies to our next check-in. Ben will check in with Timoteo vis-à-vis his schedule.
Action Items
- Clay Crosby (Unlicensed) will report what he finds regarding A) the likely timing of Superset's next release, and B) the level of effort necessary to setup an RC instance of Superset after Ben has associated every datasource with a dashboard.
- Clay Crosby (Unlicensed) will ask Jason about our CPU usage, and let us know whether he thinks our hardware can likely support 20 concurrent users.
- Clay Crosby (Unlicensed) will ask Jason to update and automate the certs associated with https://dashboard-pav.openlmis.org/superset/dashboard/6/. (It wouldn't load at all yesterday. It works better today, but Chrome's address bar still warns that it's "Not Secure.")
- Felimone Amone Junior Junior will highlight for the team all temporal coefficients applicable to the dashboard.
- Felimone Amone Junior will send the team district-level population metrics.
- Ben Leibert will discuss the timing of the upcoming documentation/training efforts, along with our associated options, with the team.