Обсуждение: Re: BUG #17784: broken URL in 15.2 release notes

On 2/9/23 9:22 AM, Andrey Lizenko wrote:
> Just in case, may be it is a CDN issue.

Just went to the URL. Works for me, and I did not touch any backend 
settings. I don't know if you were at the page prior to the cache flush.

Jonathan

On 2/9/23 10:33 AM, Andrey Lizenko wrote:
> No, its ok for now, but didn't work for a quite short time, that is why, 
> I've checked it in a fresh cacheless FF, instead of usual Chrome. My 
> suggestion about CDN were also about it - we just need to wait.

Well, on the pgweb side (the website code) we should ensure that the 
release notes are available. As part of the release SOP we do have a 
step to clear the release notes cache prior to launch. I clearly did not 
do that -- when I did my quick test it appeared to be OK, but that was 
only on the 15.2 notes.

We do have code to automatically purge portions of the cache when 
certain models update, and on inspection we don't do that when we push 
out the releases and update the versions. I'm going to work on a fix for 
that, as this has now happened a few times (and granted, is typically my 
fault) but should be easy enough to automate.

Jonathan

On 2/9/23 11:46 AM, Jonathan S. Katz wrote:
> On 2/9/23 10:33 AM, Andrey Lizenko wrote:
>> No, its ok for now, but didn't work for a quite short time, that is 
>> why, I've checked it in a fresh cacheless FF, instead of usual Chrome. 
>> My suggestion about CDN were also about it - we just need to wait.
> 
> Well, on the pgweb side (the website code) we should ensure that the 
> release notes are available. As part of the release SOP we do have a 
> step to clear the release notes cache prior to launch. I clearly did not 
> do that -- when I did my quick test it appeared to be OK, but that was 
> only on the 15.2 notes.
> 
> We do have code to automatically purge portions of the cache when 
> certain models update, and on inspection we don't do that when we push 
> out the releases and update the versions. I'm going to work on a fix for 
> that, as this has now happened a few times (and granted, is typically my 
> fault) but should be easy enough to automate.

Per a discussion on -www, we pushed a fix that should prevent (or at 
least better mitigate) the risk that this happens in the future[1].

Thanks!

Jonathan

[1] 
https://git.postgresql.org/gitweb/?p=pgweb.git;a=commit;h=ac618d1bb5f8b28ca8766e5ce6e25642f8a36539