OBIEE 12c "bad gateway" error when "printable html" is selected for a dashboard page

User_INLWQ

Hi Gurus,

I have a dashboard page with many reports (around 25). Once i select prompts values and hit apply, the reports keep running (some of them show data while others keep running).

At this point when i go to "page options" on top right corner and select Print --> printable HTML, i get "bad gateway" error.

Please suggest how to fix this as this is bothering our users for a while.

Thanks in advance.

[Deleted User]

https://community.oracle.com/tech/apps-infra/kb/articles/4-how-to-ask-questions-in-developer-community-spaces

1.) Which "12c"? There were about 30 "12c" releases. A precise and useful version would be 12.2.1.4.210620

2.) What browser and exact browser version?

3.) "Reports"? You are using 25 BI Publisher "report" objects in one dashboard page? An report and an analysis are technologically different objects and you're probably not using the right term but there is a serious difference.

4.) What do the log files say? biserver1 log? obips1 log?

If this is bothering you since a while - what investigation have you done and what have you found?

User_INLWQ

Hi Chris,

1. OBIEE 12.2.1.3.0
2. google chrome, version 93.0.4577.82 (Official Build) (64-bit)
3. Mistake, all of those are analyses.
4. Log files -- i don't see any entry in sawlog0 file of OBIPS in that timeframe, but in bi_server1.log i see below error (posting here just the first line as there is a lot of java related errors)

"java.net.SocketTimeoutException: Read timed out"

Thanks..

Gianni Ceresa

What do you expect to print if the page is still loading?

Same exact behavior if you try to print only after the page (and the content) fully loaded?

User_INLWQ

Hi Gianni,

no, when we wait till it loads fully, then no issues, happens only when tried while its "still loading".

here i am trying to understand what's behind this dashboard's behaviour, some times fully loads in 5 mins, sometimes even after 25 mins, it still keeps loading (for the same prompt values).

For ex: i ran this dashboard at 3:15 pm, by 3:22 pm fully loaded (understand that 7 mins is still not a good time)

when i ran at 4:15 pm, even at 4:40 pm, still loading.

So its not the issue of poor dashboard or RPD metadata design. Trying to figure out if some firewalls causing this or OS level parameters need to be modified.

Thanks..

Gianni Ceresa

You have a dashboard page with a large number of analysis (25), each analysis has to be executed. Depending on what else the system is doing, depending on your settings, depending on what those analysis really do, having a full loading of the page taking 7 minutes at 3pm and 25 minutes at 4pm sounds reasonable.

Not because it's 3pm or 4pm, but because you have an incredible number of moving pieces on the "what defines the full loading time of the page?" question, and all are influenced by even a larger number of parameters.

Look at all your logs, look at what your analysis are doing and how they are doing it and if you can see a pattern of where the time is spent. To have an idea of where to start investigating.

Of course having to deal with a dashboard page means you have to analyze the performance of 25 analysis, and not just every single individually but also all of them together as they can interact and help or stop each other.

Try to simplify the problem if you can identify some analysis that keeps performing more or less the same and others that change a lot. This is only a possible path to follow, there can be many more reasons related to the server RPD, the server etc.

So its not the issue of poor dashboard or RPD metadata design.

Well, that's all relative, depends if the same analysis could be executed in seconds instead of minutes because of design choices: it can still be an issue of poor modelling.

User_INLWQ

Hi Gianni,

Out of the 25, identified 8 reports which are very slow (others return data in seconds ALWAYS, these 8 reports sometimes take 5 mins, sometimes upto 25 mins), DBA identified creating indexes shall speed up those slower ones by 60 to 99%. But just want to figure out apart from having no indexes on the tables if any firewalls and server/server RPD (you mentioned this in above) is also a reason behind this performance and how to determine it.

Thanks..

just a correction .. not reports, these are analyses.

Gianni Ceresa

While your question is legit and simple to ask, the answer is fairly long and complex...

To have an idea of why this is a complex topic, there is a old series of blog posts going into a lot of details, and in this one https://www.rittmanmead.com/blog/2013/03/performance-and-obiee-test-design/ you see why an OBIEE analysis isn't as a simple thing to analyze. Not only that, but analysing the results of a performance analysis is even more challenging for multiple reasons (https://www.rittmanmead.com/blog/2013/03/performance-and-obiee-analysing-results/).

You could build the time profile of your analysis first, but don't forget the environment. OBIEE maybe generate a query that only takes 10 seconds to be executed on the database, but maybe your OBIEE query is queueing in waits 2 minutes before to be executed because the pipe connecting your BI Server to the database isn't "large" enough?

The settings of your connection pools, the database accounts used, the settings of those database accounts in the database. All that enter in the game of finding out why something takes 5 minutes now and 25 minutes in one hour.

The analysis require to know the whole environment (including the OBIEE server, the network, the DB server etc.) and to understand how OBIEE works, what happen when you execute an analysis? What are all the steps? How to measure how long does each step take?

It's complicated!

Doable, but complicated....