Posted Friday, March 15 at 5:29 PM (3 years ago)
While trying to figure out some weird access patterns on the day-job site I had the realization Pushl wasn’t actually specifying a user-agent, so it was just coming through as the generic
aiohttp one, which isn’t very friendly.
Now it sends a reasonable user-agent by default, and this can be overridden by the
--user-agent flag if you want to for your own analytics or whatever.
Oh, and I had quietly released 0.2.3 a few days ago; there were just some minor internal changes to logging and also declaring Pushl as beta, rather than alpha, software.
Posted Sunday, March 10 at 6:25 PM (3 years ago)
I’ve done a bunch more work on Pushl to try to get it more stable. In particular, I’ve made it so that it will only recurse into feeds that are on domains that were declared in the initial requests, and I seem to have cleared up some cases which were causing it to hang and also added a global timeout which will, hopefully, prevent it from hanging indefinitely.
I do wish I could figure out what is causing the hangs when they do happen though. Oh well. Some discussion of the issue below the cut.
Posted Thursday, March 7 at 10:27 PM (3 years ago)
I’ve been working on getting Pushl much more stable and reliable, particularly around a persistent “too many open files” error I was having, which turned out to be primarily due to a fd leak in the caching routines. Oops.
Anyway, there’s also seemingly a problem with how
aiohttp manages its connection pool, at least on macOS, so I’ve disabled connection keep-alive by default. However, if you still want to use keep-alive, there’s now a
--keepalive option to allow you to do that. I’m finding that it doesn’t really improve performance all that much anyway.
This is feeling beta-ready but I’ll give it a few days for other issues to shake out first.
Posted Thursday, March 7 at 12:05 AM (3 years ago)
So, I just released v0.2.0 of Pushl. It was a pretty big change, in that I pretty much rewrote all the networking stuff, and fixed some pretty ridiculous bugs with the caching implementation as well.
The main thing is now it’s using async I/O instead of thread-per-connection, so it’s way more efficient and also times out correctly.
And oh gosh, I had so many tiny but critical errors in the way caching was implemented – no wonder it kept on acting as if there was no cached state. Yeesh.
Anyway, I’ll let this run on my site for a few days and if I like what I see I’ll upgrade it to beta status on PyPI.