Since installing 9.0.1 FP4, I have had a couple of crashes where the console log indicated : PANIC: AccessAllProtected() error 107 from MapSharedRegion on pool 59 of 61 pools
I am wondering if this might be related to one of the following setings:
from notes.ini:
Using_Local_Shared_Mem=1
nsf_hooks=nlnvp
;#Trying #Editor_Memory_Scale=50-200 to get Synopsis to work better
;#Editor_Memory_Scale=200 'kills Notes eventually
Editor_Memory_Scale=140
Local_Shared_Mem_Session_ID=2
Create_R9_Databases=1
Admin_Client_Skip_Domino=1
LogStatusbar=1
DisableRoaming=1
Disable_ClientRecord=1
JavaMaxHeapSize=128MB
Debug_ThreadID=0
NSF_Buffer_Pool_Size_MB=256
Client_Clock_Server_Format=0
DisablePrinterFonts=1
Ftg_Use_Fast_Binary_Sort=1
Ftg_Enable_Gtr_Key_Check=1
Ftg_Use_Sys_Memory=1
Ftg_Disable_Paging=1
Notes_TempDir=C:\notesTemp
Debug_LS_Dump=0
OptimizeImagePasteSize=1
FulltrimFix=0
from jvm.properties:
#set max heap higher
vmarg.Xmx=-Xmx1024m
#set min heap at 64m so compression of early allocated objects occurs
vmarg.Xms=-Xms64m
#set alloction size to 512k
vmarg.Xmca=-Xmca512k
vmarg.Xquickstart=-Xquickstart
vmarg.Xnolinenumbers=-Xnolinenumbers
vmarg.Xjit=-Xjit:noResumableTrapHandler
vmarg.Dosgiext=-Dosgi.framework.extensions=com.ibm.rcp.core.logger.frameworkhook,com.ibm.rds,com.ibm.cds
#set cache size to 128M
vmarg.Dscmx=-Xscmx128m
#set maximum number of bytes in the class cache that can be used for AOT data
vmarg.Xscmaxaot=-Xscmaxaot16m
vmarg.Xminf=-Xminf0.1
vmarg.Xmn=-Xmn7m
#removed singleJVM to load only one copy of the shared classes
#removed nonFatal to let the client only start with shared classes enabled
vmarg.Dshare=-Xshareclasses:name=xpdplat_.jvm,controlDir=${prop.jvm.shareclasses.loc},groupAccess,keep
#turn off noverify
#vmarg.Xnoverify=-Xverify:none
Subject: Crash persists, PMR opened
This morning Notes crashes again with PANIC message. PMR opened with IBM.
Subject: Server Crash
For this issue, I would recommend that you open a PMR with our Domino Crash Team, and send them a copy of the NSD taken at the tiem of the crash. This is the only way to determine the root cause of the crash
Subject: Oops missing lin with PANIC message
ahum, this line was missing (from several console_mach_datetime.log files, pool numbers varying of course):
PANIC: AccessAllProtected() error 107 from MapSharedRegion on pool 59 of 61 pools
found this panic documented in Domino server crash reports, and a posssible cause was the memory allocation bumping into the 2 GB barrier. (Domino win32 on win32)
Notes is stil a 32 bits, and so I believe this to be the same issue, even when running on Windows x64.
As soon as I have another crash I wait for NSD to finish (which takes FOREVER, 15 minutes minimum, I would love pointers to speed this up!!!) and post a PMR.
Since my time and resources are extremely limited, I would love for the support team to provide me with a VM, so I can log into that, and try to recreate without my work machine having to suffer. Also, that would circumvent any setup troubles since ii don’t have admin rights on my main workhorse (which of coursesuffers most). If the VM would havve Win7/x64 and 4GB minimum (currently on a machine with 8 GB but I think the problem will show as long as the client has enough room to grow against the limit)
Subject: It’s not a server crash
It’s a Notes client crash. And I think it’s because the VM size (Virtual Size column in Russinovich’s Process Explorer) is bumping on the 2Gb limit. Which makes the Notes client hurt bad and die.
Subject: Testing possible resolution for PANIC with error 107 on MapSharedRegion
In the 8.5 foru, I found this article:
http://www-10.lotus.com/ldd/nd85forum.nsf/7756aedc25e6d81285256324005ac76c/9e8701fb0b9915b685257cdf005af274?OpenDocument http://www-10.lotus.com/ldd/nd85forum.nsf/7756aedc25e6d81285256324005ac76c/9e8701fb0b9915b685257cdf005af274?OpenDocument
8<----
I found some internal TechNote regarding this type of crash.
Basically, the crash is occurring due to an incorrect setting in the notes.ini on the server: HTTPJVMMaxHeapSize=1024M
Resolving the problem:
HTTPJVMMaxHeapSize should be set at either the new default > HTTPJVMMaxHeapSize=64M
Another option is to delete the ini parameters HTTPJVMHeapSize and HTTPJVMHeapSizeSet from the notes.ini and restart the server. These will be recreated. Check that they are set to the defaults after the restart:
HTTPJVMMaxHeapSizeSet=1
HTTPJVMMaxHeapSize=64M
If the crashs still persists, I encourage you to open a PMR to further investigation.
8<----
Since I noticed that the Virtual size of notes2.exe is over 1.9GB soon after opening my client, I’test above solution and see what happens.
Subject: Changing Notes.ini settings does not change alloc for notes2.exe
After some testing, no change to a notes.ini setting (like limiting JavaMaxHeapSize, or NSF_Buffer_Pool_Size or any of the FTG_ settings.
So I turned to the jvm.properties file.
Tuning the -Xmx
vmarg.Xmx=-Xmx768m
Finally curbed the notes2.exe Virtual Size. I’ll run a bit and report if the crashes stay away. note that this setting is now below the recommendation in:
Intermittent memory issues and hangs in Domino Designer: http://www-01.ibm.com/support/docview.wss?uid=swg21617708 http://www-01.ibm.com/support/docview.wss?uid=swg21617708
That document specs:
vmarg.Xmx=-Xmx1024m
I humbly submit that at least in my situation, this steers the Notes client too close to the 2GB abyss.
Note for any engineers trying to reproduce: on the main clientm, I also have notesHound, Ytria suite and Domnavigator installed, plus all the usual components like Sametime etc)