
IBM Lotus Notes, Domino, Domino Designer 8 Release Notes
Kernel values modified by Domino
In addition to the above, other Linux kernel value changes are needed for large scale Domino
environments. Unlike the per-process filehandle limit, which must be modified before logon of the Domino
user id, these values can be modified by Domino itself. At startup, a program named "tunekrnl" runs with
root authority and sets these values in the /proc filesystem. Once set, these /proc values will not change
until Linux is (re)booted, at which time they revert to their default values. The exception to this is
"mapped
base", which is a process-local setting made at each Domino startup. The "tunekrnl" log of
/proc filesystem changes tends to vary from many (following a Linux boot) to few (subsequent Domino
startups). The kernel values modified by "tunekrnl" are:
kernel parameter
default value tunekrnl value description
/proc/sys/fs/file-max (as low as 8192) 131072 maximum number of
system-wide file descriptors
/proc/sys/kernel/sem 250 32000 32 128 250 256000 32
1024
shared semaphore allocation
sizes and limits
/proc/sys/kernel/shmmax 32 M 256 M maximum shared-memory
segment size
/proc/sys/net/ipv4/tcp
fin
timeout 60 15 retransmission timer for a FIN
(socket close) in seconds
/proc/sys/net/ipv4/tcp
max
syn
backlog 1024 16384 maximum value of the listen()
connection backlog
/proc/sys/net/ipv4/tcp
tw
reuse 0 1 1 enables "time-wait
assassination" allowing for
rapid re-use of sockets
/proc/sys/net/ipv4/ip
local
port
range 32768 61000 1024 65535 increases the range of port
values available for connect()
/proc/<pid>/mapped
base 1 Gigabyte 16 Megabytes "floor" address value above
which memory allocations may
occur on SLES
Swap configuration for Linux for System z
SLES8 (31-bit) supports up to 2 Gigabytes of memory; a Domino server supporting thousands of users
will typically require more than 2 Gigabytes of memory, causing swapping to occur in this environment. At
least 1 Gigabyte should be allocated as a swap "disk", however actually swapping to disk (DASD) hinders
performance. One mechanism to keep swapping within system memory is XPRAM. The following
general steps are an example of the configuration of 1 Gigabyte of XPRAM for Linux for System z in a
native environment:
1.
Configure at least 1 gigabyte extended memory from the hardware console. Linux for System z will
automatically detect that.
2.
Configure the xpram driver.
3.
The following Linux commands set swap space 1G from a 3G partition that was made into a file
system and mounted off of the xpram
fs directory:
insmod xpram devs=1 sizes=2000000
mkswap /dev/slram0
swapon /dev/slram
91
Comentários a estes Manuais