If anyone is wondering how it’s going, RSP services have now shipped 4 BILLION rows of catalog results since we opened the Early DP2 doors on Monday. Query success rate was 99.91%, and we’re planning to make a deep dive post on how to set up your query for success in the near future.
And now from this week’s Patch Thursday annals:
Faster queries!
We do a bunch of load testing before a release, but for sure nothing beats the real thing, so we’re always poring over the metrics in a release week to see what y’all are actually doing. And that’s how we saw a perfectly sensible query clock at 14 hours 
In not unrelated news, we have now applied indices to the dp2.SSObject.designation and dp2.mpc_orbits.designation columns, so at least one user can now enjoy the new under-4-second query time for that join.
Questions along the lines of “why is my query taking a seemingly unreasonably long time” are welcome as support topics. We’ll learn something, or you’ll learn something, or both.
Speaking of slow queries…
There’s a system behaviour right now that if you’re (quite) unlucky your query can sometimes end up with a 50-second lag to however long it would have taken for you to get results, while most times it’s just fine. We understand why this is happening and are working on a fix.
This is the kind of situation where IRL some of us (cough) would launch into a long explanation as to what’s happening only to watch our interlocutor’s soul slowly leaving their body. But this is skippable text, so, quick poll: In such a situation, should we…
- Explain it all, right here, right now!
- Explain sometimes, but maybe in a different (non-news) post
- Eek no, we don’t need long technical explanations, just the fixes ma’am
0 voters
Fewer failures!
RCW attendees may have heard my strangled cry of horror (sorry…) as a notebook cell turned red on the giant screen during a demo. We identified the resource starvation responsible and have now showered the relevant operation with gifts of memory and CPU so that’s dealt with.
Folks have been lovely with “it’s fine maybe the system is just busy” sentiments this week, and we appreciate your goodwill 
, but this is a system that is built to still work when it’s busy! So do report these issues via the normal support mechanisms so we can squash more bugs before the DR1 firehose comes to us all.
Black cat in a dark cave at night?
We’ve seen some cases of people thinking their notebook terminals have hung - in fact they have gotten in a mode where there is black text on black background (or white text on white). Vexingly this seems to be a JupyterLab bug that only triggers with a narrow range browser/OS combinations.
We’re hoping for a fix from upstream but in the meanwhile if you see this, just pick a different theme (light or dark) instead of asking it to follow whatever your system is doing.
DP1 or DP2?
Now that there’s a choice of data release, we have moved the obs_collection column in portal results ahead of the queue so it’s obvious what you’re getting.
Remember you can also use the dropdown in the header to further filter your results.