Quicksilver task times out with no notification about quicksilver-examples

[joshkoenig]

Currently quicksilver operations are limited to a 120 second timeout. If you're not getting results, that's likely the reason why. I think it's far to say we should make that clear, which I'll check up on.

More generally, you should be able to trigger this same operation (with no timeouts) as a standard part of the database clone workflow within the UI.

[gmcinnes]

Hi @joshkoenig . I'm going to enumerate the issues for brevity here. Don't read anything else into my terseness.

WRT the use case for this:

• I know I should be able to trigger this using the UI, but for some reason it doesn't work on all my sites
• One site in particular has issues with the urls of the main navigation not being replaced, which makes it a huge pain to remember fix with `terminus wp "search-replace foo bar" every time I make a multidev environment
• In particular I have a ticket open with pantheon support since June 2015 about this that wasn't really resolved. I didn't push the issue.
• Also, it would be handy to force this to happen automatically every time the database is cloned from live.

As for the possible timeout:

• My db is bigger than average (174MB in the backups). It does take a bit of time to do a wp search-replace locally, but the db is mostly full of audit logs that my client doesn't want to lose, so I can't do much about it.
• Can someone confirm that's the issue?
• If so, I agree, you should document it
• There should be some sort of output in the workflow logs in this failure case. As I noted there's not only no record of the workflow failing, there's no record of it even starting.
• As well, any debugging statements, even ones added before the potentially problematic, long running code, also don't make it to the workflow log, which is certainly violating the PLOS.

I'm not sure this is the right place to discuss this, 'cause it doesn't really seem like an issue related to a "quicksilver-example". Let me know if you'd prefer to move this off thread.

[joshkoenig]

I know I should be able to trigger this using the UI, but for some reason it doesn't work on all my sites.

By any chance do the sites where this doesn't work have no WordPress icon on your home screen? If they aren't properly associated with WP as their "framework" then the WP-specific dashboard options will not be present.

It does take a bit of time to do a wp search-replace locally.

If it takes over 120 seconds, this is very likely the issue. This is documented, but it's easy to miss. I'll follow up with our support team to dig in and make sure this is the case. Clearly room for improvement here.

[gmcinnes]

By any chance do the sites where this doesn't work have no WordPress icon on your home screen? If they aren't properly associated with WP as their "framework" then the WP-specific dashboard options will not be present.

No. The options are there. They just don't work. Some urls are replaced, some are not. When I run a search-replace with terminus it replaces the rest of them and fixes it.

If it takes over 120 seconds, this is very likely the issue. This is documented, but it's easy to miss. I'll follow up with our support team to dig in and make sure this is the case. Clearly room for improvement here.

FWIW, on my local dev machine that looks like:

time wp search-replace https://www.marsdd.com http://marsdd.dev:8888

.....


Success: Made 113123 replacements.

real    1m33.815s
user    0m0.817s
sys 0m0.126s

so, less than 120 seconds, but I have no way of knowing how long it would take inside a Pantheon container.

[joshkoenig]

No. The options are there. They just don't work. Some urls are replaced, some are not. When I run a search-replace with terminus it replaces the rest of them and fixes it.

Well that's weird. They do the exact same thing on the back-end. I'll look up the case number.

so, less than 120 seconds, but I have no way of knowing how long it would take inside a Pantheon container.

Likely a bit longer if only because of the remote DB connection. That adds a few milisecond penalty to each request vs when it's all on one machine. if you're doing 100k replacements, that adds up. We've sponsored some improvements to the performance of WP-CLI that should help speed things up in general, but they're only rolling out now.

We've identified the need to get better feedback on timeouts and should be able to improve this situation in the coming weeks. I'll follow up on your support tickets with specific confirmation that this was the cause, as well as looking into why you're not able to use the time-unlimited built-in dashboard function to take care of your search-and-replacing.

[joshkoenig]

This is now partially fixed. We are handling timeouts properly internally.

Still work to do in order to get the timeouts to be presented nicely in the Terminus workflows --watch output.