Installation guide

Release Notes
Optionally increase the max number of processID values
The kernel defaults to using a maximum of 32,768 processID values. Scyld ClusterWare automatically increases this default
to 98,304 [= 3*32768], which likely is adequate for small- to medium-size clusters and which keeps pid values at a familiar
5-column width maximum. Because BProc manages a common process space across the cluster, even the increase to 98,304
may be insufficient for very large clusters and/or workloads that create large numbers of concurrent processes. The cluster
administrator can increase the value further by using the sysctl command, e.g.,
/sbin/sysctl -w kernel.pid_max=N
directs the kernel to use pid values up to N. The kernel (and BProc) supports an upperbound of 4,194,304 [= (4*1024*1024)].
To set a value N that persists across master node reboots, add an entry
kernel.pid_max=N
to /etc/sysctl.conf.
Optionally increase the max number of open files
The kernel defaults to a maximum of 1024 concurrently open files. This value may be too low for large clusters. The cluster
administrator can add a nofile override entry to /etc/security/limits.conf to specify a larger value. Caution: for
nofile, use only a numeric upperbound value, never unlimited, as that will result in being unable to login.
Issues with Ganglia
The Ganglia cluster monitoring tool may fail for large clusters. If the /var/log/httpd/error_log shows a fatal error of
the form PHP Fatal error: Allowed memory size of 8388608 bytes exhausted, then edit the file /etc/php.ini to increase
the memory_limit parameter. The default is memory_limit = 8M can be safely doubled and re-doubled until the error goes
away.
Post-Installation Release of Updated Packages
From time to time, Penguin Computing may release updated Scyld ClusterWare 4.9 rpms to track Red Hat kernel security or
bug fix errata or to fix critical Scyld ClusterWare problems. You can check for the availability of updated Scyld ClusterWare
rpms by doing:
yum list updates --disablerepo=* --enablerepo=cw*
If updates are available, you should first download the latest version of the Scyld ClusterWare 4 Release Notes from the
Penguin Computing Support Portal (http://www.penguincomputing.com/files/scyld-docs/CW4/ReleaseNotes.pdf) to ensure
you have the latest guidance before updating your cluster. In general, if you choose to update Scyld ClusterWare packages,
then you should update all available packages.
If your cluster uses Panasas storage, then before updating Scyld ClusterWare you must ensure that a Panasas kernel module
is available that matches the Scyld ClusterWare kernel that will be installed. See the section called Important for clusters
using Panasas storage in the About This Release introduction for more information.
10