Question: errors from queries involving pg_catalog tables running while things are being dropped about sqlalchemy-redshift

[jklukas]

The cache lookup failed error is definitely familiar to me, as we occasionally see it during some of our nightly maintenance tasks where we're recreating views.

Maybe these queries shouldn't be reflecting the entire schema at all, and we should limit them to looking only at individual tables if the current broad approach can cause errors and isn't actually allowing us to reuse the reflection cache for multiple tables.

[atrigent]

I think it is allowing reuse of the results of the reflection query, as discussed in #101. However, yes, it is sometimes causing errors. We think that restricting the queries to the correct schema will resolve our issue, but there may be cases where even that isn't enough.

[willnewby]

We've also had this issue. Any idea on merge timing for the above PR?

[jklukas]

Looks like there were some lingering issues in #105 that never got resolved. Do you have any thoughts on how those could be addressed?

[grantatspothero]

Still seeing error 1 in 2018/almost 2019. The problem goes away when a background job that drops tables is completed.

@jklukas any update on PR #105 it looks like it was never merged?

[jklukas]

I haven't been working with Redshift much in the past year, so it's difficult for me to jump back in and spin up the context to understand whether #105 is safe to merge or not. It's probably worth starting from scratch at this point.

It looks like we'll be merging #147 imminently, which is a simpler change to reflection which may or may not help with this issue. I'd be very open to someone evaluating whether #147 has a bearing here and if needed opening a fresh PR that absorbs the proposal and discussion in #105.

[cscetbon]

Still seeing this in 0.7.9 .... Any idea of a workaround in the meantime ? Those issues have been opened for years now