Hi,
We had an issue where our SAP ECC system was crashing due to LWP value reaching more than 56K.
The Unix team found from dtrace logs that the segkpsize had reached limit and hence the system was crashing. As a solution, the unix team increased the value of segkpsize : 524288 and now the lwp upper limit is around 128K. The segkpsize was earlier set to segkpsize: 262144.
4. However now that we are tracking the lwps count on the CI-DB server, we are observing that between 12:00 and 20:30, the value of lwp's peaks upto 70K and we are concerned that it would kep increasing.
5. On checking with ps -leafo nlwp,user,pid,args | sort -nk 1 command we are seeing around 175 to 200 oracle processes running with 258 lwp units each.
We have some Queries and concerns to investigate incase this issue reoccurs or the number of lwp increases.
1. Is there any normal acceptable value for lwp count in heavly used SAP system. Our load at the moment is
prstat 2 5 | grep lwp
Total: 698 processes, 54951 lwps, load averages: 21.83, 25.52, 24.82
Total: 698 processes, 54951 lwps, load averages: 21.83, 25.50, 24.82
Total: 698 processes, 54951 lwps, load averages: 21.82, 25.47, 24.81
Total: 698 processes, 54951 lwps, load averages: 21.87, 25.45, 24.80
Total: 698 processes, 54963 lwps, load averages: 21.94, 25.44, 24.80
2. what are the factors in SAP that control the lwps count.
3. We can identify the process running with high number of lwp is there any way to relate them to SAP process and reduce the lwps consumption.
4. is this a Solaris 10 limitation.
Any help would be appreciated.
Thanks
Ahmad Faraz