Hi,
We have encountered a situation like the below on some transactions that make it look like 5 seconds was spent doing a web request from an apache web server to an application backend server but according to the "Elapsed time" this only goes up until 5ms and the "Exec Total" time is also not adding up to this. This kind of gives a false positive in that it looks like the Apache web server is the cause for the slow response of the transaction. Are we interpreting this incorrectly?
If anyone wants I can send them the actual exported purepath to look at it.
From the transnational flow it also looks like 86% of the time was spent on the web server tier?
I have attached the call stack tree.Call Stack Tree.xlsx
Kind Regards
Richard
Answer by Andreas G. ·
Hi Richard
Any chance you can also export the PurePath? Makes it a little easier to read - or - maybe attach a screenshot of the PurePath so you can blank out information you dont want us to see.
If I look at the Excel file it seems that most of the time is "spent" when handing the request from Apache to Java. The Elapsed time indicates that the JVM doesnt get to handle that reqeust until 4,08s. Typically this can be explained if the JVM cannot accept the incoming requests fast enough, e.g: running out of worker threads. Therfore the request is put in a queue and picked up at a later time. that woudl explain it
I would look at the Threads in Java -> you can see that in the Process Health Dashboard.
Another thought: are these two servers running on the same or different physical machine? If they run on a different machine you may also have some delaying factors between these two containers, e.g: network constraints or a components such as a firewall/proxy between these two tiers.
Andi
Thanks for the response Andreas if this is the case it makes perfect sense but the elapsed time is displaying milliseconds though not seconds thats why its just a bit confusing to me?
I will e-mail you the purepath
Kind Regards
Richard
JANUARY 15, 3:00 PM GMT / 10:00 AM ET