aldorexbraam

Coresight 2012, plotvalues and server load

Discussion created by aldorexbraam on Sep 3, 2014
Latest reply on Sep 5, 2014 by Gregor

Today while troubleshooting a performance issue at one of our customers sites with PI Server/Coresight i stumbled accross a puzzling issue. Insiights of the community are appreciated.

 

Customer had the issue that the archive subsystem of the server would occasionally stop handling calls, The service would remain running but the message log would show exceptions like

 
T:4448 PT(ptid=666,recno=666,loc=5,flag=4): ptwait tick/timeout=27000/270000, RID|LLL|LC|LI|NAL=0|2|2|0|253 Lock[253].uc=5 QL.sc/tid=0/4448 threads=2,0,0

it turned out the each time (alltough many other tags were queried it would choke on two specifiec tags (pointid marked orange), which happened to contain archive data for roughly every second. 

 

Observed was while monitoring the avg disk queue length that the server clearly had disk IO issue, thus leading to request time-outs (marked green)...So in this case Coresight was merely making a hardware issue with the PI server more apparent 

 

So far all clear......

 

Now comes the puzzling...

 

While trying the mimic the load strain using PI procesbook we observed a dramatic server load difference in favor of PI processbook...thinking of this that made sense since PI PB uses the PlotValues() SDK call for its trending...

 

This leads me to the path where I suspect that Coresight 2012 ( and possibly 2013) is NOT using that function for its data retrieval...

 

Can anybody

 

a) confirm/debunk  this to be the case ?  

 

b) Share similar experiences on Coresight causing high server strain/ archive walks when trending data ? 

 

Regards, 

 

Aldo Braam 

 


 


Outcomes