Timeout on deleting large number of packages from PPA

In https://launchpad.net/~mysql-ubuntu/+archive/ubuntu/mysql-5.7/+delete-packages, I selected ~70 packages for deletion and get a timeout. This seems consistently repeatable. As a workaround, deleting smaller numbers at a time appears to work, but is painful because they don't disappear immediately so it is awkward to identify subsequent chunks.

OOPS-7f2d9fc214a7d52b3b85ed006d09e230

13:07 <cjwatson> I suppose it's a bug, but you may just have to divide it into smaller chunks. The OOPS isn't full of repeated queries or anything.

13:08 <cjwatson> Oh, there's like 5 seconds of non-SQL time.

13:10 <cjwatson> rbasak: Is this repeatable?

13:10 <rbasak> cjwatson: yes

13:10 <rbasak> I retried a few times.

13:10 <rbasak> I can try in small chunks I suppose.

13:11 <cjwatson> There's no obvious reason it should be spending several seconds contemplating its navel outside an SQL query there ...

13:11 <rbasak> I can leave it for examination if you wish. No real hurry.

13:11 <cjwatson> I mean there are quite a lot of BinaryPackagePublishingHistory ids but still.

13:11 <rbasak> I was just cleaning up as requested :)

13:12 <cjwatson> We probably won't have time to get to it very soon, and we have the OOPS, so go ahead.

13:13 <rbasak> OK

13:13 <cjwatson> I suppose it *might* just be spending a long time materialising four rows for the sake of a single ID.

13:13 <cjwatson> 4 * len(bpph_ids) I mean

13:14 <rbasak> Since they don't disappear straight away, it's a little awkward for me to do a chunk at a time. I'll do it over a day or two I guess.

13:17 <cjwatson> We could probably simplify the query there, since we don't care about being able to order the BPPHs by bits of the other rows.

13:18 <cjwatson> rbasak: Please file a bug and mention the OOPS ID above.

13:18 <rbasak> OK will do