Closed sgreenstreet closed 6 years ago
On Ira's suggestion, I went direct to the NEOexchange (and so bypassed the Varnish cache and it's potential timeout) site using http://docknode06:8200/neoexchange/ephemeris/?target=2015+SZ2&utc_date=2015-10-01&site_code=Q63&alt_limit=30.0 to directly compute the ephemeris and got the same 502 Bad Gateway.
My suspicion is that the very close pass of this object is causing the perturbing routine in SLALIB
to err, become perturbed, and either take a really long time or do something wacky. As a short term fix, I will look at increasing the timeout in the nginx
config (Edward, if you want to step in and do it and then assign it back to me). Then we can at least see the output and investigate further.
I think this is a timeout. I just grabbed the latest DB and ran the code locally. That ephemeris page was a bit slow in loading but eventually displayed. I'm pulling down the docker image to try out locally. I think it might be running out of memory. I'll test it locally before changing the setting in neox/docker-compose.yml
.
I couldn't test this with a local container because communicating with db01sba
in real time from here caused more problems.
I'm going to assign this to you Tim. Increasing the memory allocation on the (live) docker container to 512Mb didn't have an effect either.
Closing. We know have orbit integration via find_orb
for close passing objects which gives much better results than the perturbation scheme
The calculate ephemerides computation led to a "Bad Gateway" page, possibly due to timing out before finishing the computation. The schedule observations link had a difference of 15 arcmin in RA and 3.5 degrees in Dec between that reported here and that on the MPC page. The window of observation midpoint was also computed as a time when the object was only at +2 degrees altitude.