Automic Workload Automation

What is the design behind searching objects from the 'Search' window on top right corner in AWI? 
Does the objects get cached on automation engine start? 
We are on 12.0.1, I understand JWP helps with searching for the objects in this version. 

The JWP builds a cache of objects and attempts to keep it current. The cache is persistent across engine restarts. The exact mechanism, e.g. whether the cache is rebuilt on a schedule, or whether there is a message-based cache invalidation, is not known, at least not to me.

I had at least one case where we had to wipe the cache and had to have it rebuilt, because it kept referencing items that had already been fully deleted from the system.

If anyone from Automic would feel compelled to fill in the blanks with e.g. am explaination on how the cache mechanism is actually implemented, I'd be very curious as well.

Cheers,
Carsten

Carsten_Schmitz_7883 thanks for your response. How did you wipe the cache? 

Am working on mass updating the object names via backend. The new names are reflected on the thick client search. However, AWI search still reflects the old names of the object. I tried restarting JWP/Automation Engine, AWI Tomcat but not luck. 

Kesavan Elangovan said:

Carsten_Schmitz_7883 thanks for your response. How did you wipe the cache?

The directory where your AE is installed has a sub-folder like this:

bin/indexCache-12.1

(or -12.0 or somesuch).

That's the JWP cache, afaik. You can remove the entire directory, it will get re-created after a short time (but best practice, don't delete it but move it somewhere else, just in case).

If your cache has stale entries that are not fixable any other way, I also encourage you to open an incident and submit your stale cache folder, so Automic can possibly fix the indexing.

hth,
Carsten

Carsten_Schmitz_7883
 Thanks, I was able to completely delete the indexCache folder and restarted the JWPs to reflect the changes. 

I'm still on v11, but if I read this thread correctly, the v12 search tool no longer searches the actual system, but a stale cache view of the system.  Am I the only one that thinks this was a bad decision?  Or at least they should provide a 'clear cache' button?

I can think of use-cases where someone will make changes and immediately go into search to validate those changes....

petwir
I wouldn't say the searches are done on the cache. It seems cache is constantly getting updated based on updates to the objects from Thick Client or AWI. 

Our scenario is, we have moved number of jobs from a legacy scheduling system to Automic. 
Now we have decided to update the job names/title by exporting the data into an excel sheet and forming SQL update queries. These updates did not reflect in the search, that's the reason I was exploring ways to clear/rebuild the cache. 

Pete Wirfs said:

I'm still on v11, but if I read this thread correctly, the v12 search tool no longer searches the actual system, but a stale cache view of the system.  Am I the only one that thinks this was a bad decision?

Tough one. A cache is a good thing if done right. And to be honest, I (or anyone else outside of Automic PM and the dev who wrote it) probably don't know how exactly it is implemented.

E.g. it could be possible that each event that leads to a stale cache entry (e.g. deleting objects) invalidates just that cache entry instantly (ofc with major drawbacks to cache performance). I don't know if it is designed that way, and if they currently do something to that effect, they're probably just initiating a re-indexing, which isn't an atomic operation. Plus, I have reason to believe it does currently not work right in all circumstances (e.g. my case about it listing deleted objects for weeks). I have called upon Automic in another thread to explain the exact implementation details to us. However, I'm not exactly holding my breath.

So yeah, at this time, and with the added fact that Automic apparently didn't test this systematically, and their support appears to have been ordered to ignore that which they can not replicate - yes, it's at present a rather bad design decision.

A "clear cache" button is just a band-aid onto that (one could probably build an Automic job to do that, you can probably even leave the JWP running while deleting the folder).

And this will not be the only time we'll be discussing caches and friends. There's also a Tomcat cache and browser caches, and my company (and many others) enforces the use of a proxy server, which might also cache (and besides, also may introduce funky effects and make debugging harder even with no cache).

Isn't this whole browser-based thing fun? :)

more food for my AWI aversion  >:)

At the moment I feel a bit overeaten.....

Wolfgang Brueckler said:

At the moment I feel a bit overeaten.....

Ja, ich kann auch gar nicht so viel klicken, wie ich reporten möchte :pensive:

(sorry, this one just doesn't work in English at all ... we'll be returning to our regular, english language inspired programming in just a moment)

Anyway I'll try it this way...

Carstens'  attempts of clicking and reporting will soon reach the ballpark of googolplex ...

cheers, Wolfgang

PS: Now you may guess whats the source of google's name :-)

FrankMuffke I'm sorry, you didn't quite catch the "hidden" meaning that was a direct reference to your mentioning of "overeating".

It works on two levels, but possibly the underlying proverb is not that well-known in Austria. After all, you have used many words of Austro-German dialect I had no clue they existed, so ... it probably works both ways :)

Carsten_Schmitz_7883
Ahhh now I can imagine the hidden meaning of your reference to my posting...
THX for the explanation :-)

I 'll attend a spaecial training class - German hidden meanings in combination & context of Austrian dialects.
Short code: GHMCCAD

cheers, Wolfgang

PS: yeah sure this works in both directions!

Just to add a little more info. The backend for automic globale search is Apache Lucene. Which is a widely accepted, solid search engine framework. So in general I wouldn't call it a bad design decision as long as it's well implemented.

Having Lucene one can use Luke to display and modify Lucene indexes. Might be an option for advanced debugging.

Matthias