Details
-
Bug
-
Resolution: Unresolved
-
Minor
-
None
-
None
-
None
-
2
-
9223372036854775807
Description
I'm trying to find the proper way to adjust the ptlrpc thread pool. The documentation for the ptlrpc module seems to be out of date and doesn't match what's in the current module. The two module parameters I find in the documentation (max_ptlrpcds and ptlrpcd_bind_policy) both appear to be "obsolete" and the other pertinent module parameters are undocumented. Here's what I see in the current ptlrpc module:
filename: /lib/modules/3.10.0-327.el7.x86_64/extra/kernel/fs/lustre/ptlrpc.ko
license: GPL
version: 2.8.0
description: Lustre Request Processor and Lock Management
author: OpenSFS, Inc. <http://www.lustre.org/>
rhelversion: 7.2
srcversion: 5854A702057A87D76A45AC3
depends: obdclass,libcfs,lnet
vermagic: 3.10.0-327.el7.x86_64 SMP mod_unload modversions
parm: enc_pool_max_memory_mb:Encoding pool max memory (MB), 1/8 of total physical memory by default (int)
parm: max_ptlrpcds:Max ptlrpcd thread count to be started. (int)
parm: ptlrpcd_bind_policy:Ptlrpcd threads binding mode (obsolete). (int)
parm: ptlrpcd_per_cpt_max:Max ptlrpcd thread count to be started per cpt. (int)
parm: ptlrpcd_partner_group_size:Number of ptlrpcd threads in a partner group. (int)
parm: ptlrpcd_cpts:CPU partitions ptlrpcd threads should run in (charp)
parm: suppress_pings:Suppress pings (int)
parm: test_req_buffer_pressure:set non-zero to put pressure on request buffer pools (int)
parm: at_min:Adaptive timeout minimum (sec) (int)
parm: at_max:Adaptive timeout maximum (sec) (int)
parm: at_history:Adaptive timeouts remember the slowest event that took place within this period (sec) (int)
parm: at_early_margin:How soon before an RPC deadline to send an early reply (int)
parm: at_extra:How much extra time to give with each early reply (int)
parm: ldlm_num_threads:number of DLM service threads to start (int)
parm: ldlm_cpts:CPU partitions ldlm threads should run on (charp)
parm: ldlm_enqueue_min:lock enqueue timeout minimum (uint)