cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

About the "[BufferMaximizingServerMemoryTracker] Memory corridor upper limit [85.0%] violated [used = 85%], removing PurePaths that have already finished analysis" log

khanada
Contributor

Hi


Please tell me about the subject log.


Immediately after the target Info log was output, "PurePath discarded early" occurred in large numbers.


Reference log

2020-03-02 12:29:33 INFO [BufferMaximizingServerMemoryTracker] Memory corridor upper limit [85.0%] violated [used=85%], removing PurePaths that have already finished analysis

2020-03-02 12:30:34 INFO [BufferMaximizingServerMemoryTracker] Memory corridor upper limit [85.0%] violated [used=97%], removing PurePaths that have already finished analysis

2020-03-02 12:59:45 INFO [BufferMaximizingServerMemoryTracker] Memory corridor upper limit [85.0%] violated [used=87%], removing PurePaths that have already finished analysis


①Does AppMon have a memory release mechanism other than the Java standard GC?


②Do you know specifically what the target threshold is for memory?


4 REPLIES 4

First question I have, do you have proper scaling according to amount od agents and traffic on your environment according to doc?

https://www.dynatrace.com/support/doc/appmon/administration/server-administration/sizing-settings/

According to my experience in Appmon server there is just regular GC. According to your log, threshold is 85%, but server can allocate more if needed. This may be mechanism that is preventing server freeze because of 100% suspension.

Sebastian

Hi Sebastian,


Thank you for your reply.


Yes, I will guide customers to sizing.


According to the customer's network team, the corresponding time period was the same transaction volume as normal time.


However, looking at the attached image, the number of PurePaths seems to be increasing rapidly.


PurePath needed to be recorded at 12:30.

However, due to the interruption of AppMon's FullGC, PurePath is destroyed and recorded at 12:31.


Therefore, there is a difference between the amount of transactions that the network team has confirmed from the viewpoint of infrastructure and the amount of transactions that AppMon checks.


Do you think my guess is right?

Is there any chance that you have batch jobs that are executed despite normal traffic? This may be root cause of spike. Discarted purepaths are lost, they are not analysed again according to my knowledge (but I'm not 100% sure here) .

Sebastian

Hi Sebastian,


Thank you for your reply.


As far as I know, batch jobs are not running on the target system.


We haven't seen the session data yet, but there is probably a big difference between measuring transaction-based and agent-based Web requests.


We do not believe that the number of transactions has actually increased.


It is necessary to process "Record of PurePath that has not progressed" and "Record of PurePath of normal transaction" by FullGC.


For the above reasons, the memory usage will be higher than usual.


As a result, I think that AppMon's own memory release processing was executed.

(Delete PurePath in memory)


Therefore, the cause is insufficient memory of the AppMon management server.