Lotus Domino: Verificando o NSD: Difference between revisions

From Wiki
 
(One intermediate revision by the same user not shown)
Line 19: Line 19:




=== OSLockSpin(??) ===
=== OSLockSpin ===
When investigating the NSD server thread stacks, a specific stack always showed to be the most CPU consuming thread. Using OS tools, the thread always calls OSLockSpin().
When investigating the NSD server thread stacks, a specific stack always showed to be the most CPU consuming thread. Using OS tools, the thread always calls OSLockSpin().


Line 41: Line 41:
  [29] 0x090000000049f4f4  _pthread_body(??) + 0xdc
  [29] 0x090000000049f4f4  _pthread_body(??) + 0xdc
  ###################################
  ###################################


=== TCPEndp_CheckSocketReady ===
=== TCPEndp_CheckSocketReady ===
Line 55: Line 54:
  [4] 0x09000000023cd794  cmd_poll(0x151800001edc, 0x112f2b208, 0x112f2b200, 0x112f2b128) + 0x258
  [4] 0x09000000023cd794  cmd_poll(0x151800001edc, 0x112f2b208, 0x112f2b200, 0x112f2b128) + 0x258
  [5] 0x0900000000b11314  nti_poll(0x1edc00001edc, 0x112f2b208, 0x112f2b200) + 0x214
  [5] 0x0900000000b11314  nti_poll(0x1edc00001edc, 0x112f2b208, 0x112f2b200) + 0x214
= Ver também =
* [[Lotus Domino e Notes|  Mais artigos sobre IBM Domino e Notes]]
[[Category: IBM Domino]]
[[Category: Lotus Domino]]

Latest revision as of 15:01, 27 February 2014

Esse artigo lista alguns itens a observar na hora de analisar o nsd


Indice

MailExpandNamesExt

function MailExpandNamesExt was used to expand all groups in a recipients list. From this stack, you can tell that the delivery threads are busy expanding groups in the recipient's list.

Checking in mail.box, the pending mails are mass mails with large groups in bcc field.

###################################
# thread 6/55 :: router, pid=xxxx, tid=xxxx #
###################################
...
[8] 0xd1c2ee98 MailExpandNamesExt(??, ??, ??, ??, ??, ??, ??, ??) + 0x340
...

See this technote


OSLockSpin

When investigating the NSD server thread stacks, a specific stack always showed to be the most CPU consuming thread. Using OS tools, the thread always calls OSLockSpin().

OSLockSpin(??) is a low level function in Domino code used to grab a mutex from the OS to prevent other threads from doing the same operation at the same time.

###################################
###### thread 9/49 :: router, pid=581714, tid=3948649, ptid=9510) ######
###################################
[1] 0x090000000011597c  __fd_select(??, ??, ??, ??, ??) + 0xac
[2] 0x0900000000846b80  select(0x0, 0x0, 0x0, 0x0, 0x119879420) + 0x28
[3] 0x0900000000846b14  unix_usleep(0x157c0000157c0) + 0x6c
[4] 0x09000000006c6450  OSDelayThread(0x5800000058) + 0x48
[5] 0x0900000000699cac  OSLockSpin(0x700000004ada954) + 0x1b4
[6] 0x090000000069b1cc  OSUnlockReadSem(0x700000004ada910) + 0xd4
[7] 0x090000000165b178  CRWSEMReadLock::~CRWSEMReadLock()(0x119879710, 0x200000002, 0x0) + 0x3c
[8] 0x0900000001658f88  CDirAttributeTypeDescriptions::GetNext()(0x119879ab0) + 0x70
[9] 0x0900000001658d7c  CDirAttributeTypeDescriptions::AssertAll()(0x119879ab0) + 0x3c
 ...
[27] 0x0000000100067aa8  TransferThread(0x1) + 0x218
[28] 0x09000000006c2a80  ThreadWrapper(0x0) + 0x118
[29] 0x090000000049f4f4  _pthread_body(??) + 0xdc
###################################

TCPEndp_CheckSocketReady

Possible network bottleneck, in http/router/server processes when network I/O waiting for input to come while handling HTTP or NRPC request

###################################
###### thread 30/115 :: server, pid=327930, tid=1904809, ptid=2315) ######
###################################
[1] 0x090000000011597c  __fd_select(??, ??, ??, ??, ??) + 0xac
[2] 0x09000000023cdec4  select(0x1600000016, 0x112f24e50, 0x112f26e50, 0x112f28e50, 0x112f24e28) + 0x28
[3] 0x09000000023cca54  TCPEndp_CheckSocketReady(0x1000000000001, 0x112f2b208, 0x1f4000001f4) + 0x6dc
[4] 0x09000000023cd794  cmd_poll(0x151800001edc, 0x112f2b208, 0x112f2b200, 0x112f2b128) + 0x258
[5] 0x0900000000b11314  nti_poll(0x1edc00001edc, 0x112f2b208, 0x112f2b200) + 0x214


Ver também