summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
* Make waitfor set LANG=C before running expect.Martin Schwenke2011-01-251-0/+1
| | | | | | | | | | | | For some reason when creating a base image on RHEL6.0, expect exits before it finds the designated pattern. Using strace or expect -d makes things mysteriously work, so debugging is very difficult. On a whim, I wondered if expect is getting confused about multibyte characters, so tried setting LANG=C. This either fixes or works around the problem. It will do for now... Signed-off-by: Martin Schwenke <martin@meltin.net>
* Make setting of MEM happen in post_config_hooks() by default.Martin Schwenke2011-01-181-1/+15
| | | | | | | | | The previous default of 262144KB is not enough for RHEL6. We use the old default for RHEL5 and set an intial default of 512000KB for RHEL6. A smaller limit may be possible for RHEL6... but this works. :-) Signed-off-by: Martin Schwenke <martin@meltin.net>
* $ISO and $SYSTEM_DISK_MOUNTS now automatically set via post_config_hooks.Martin Schwenke2011-01-181-2/+24
| | | | | | | | | | | New function system_disk_mounts() returns a default value for SYSTEM_DISK_MOUNTS that depend on $RHEL_VERSION. New function rhel_post_config_hook() resets SYSTEM_DISK_MOUNTS and ISO if they are equal to the magic value "@uto". This is now the default value for both of these variables. Signed-off-by: Martin Schwenke <martin@meltin.net>
* New hook post_config_hooks.Martin Schwenke2011-01-181-2/+8
| | | | | | | | | | | | | | | | This is run after all configuration has been processed. That is, after command-line options processing. It is also run before other exits where you can reasonably expect configuration to be complete (--dump, -e <expr>). It is not run before -E. If you want to run in combination with -E then add "run_hooks post_config_hooks" to the expression (or to a previous -E option). However, bear in mind that this may preempt any magic the hooks might otherwise do after subsequent configuration option processing. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Change handling of packages in kickstart with new variable RHEL_PACKAGES.Martin Schwenke2011-01-185-202/+165
| | | | | | | | | | | | | $RHEL_PACKAGES points to a file containing the standard package list, which is inserted into the kickstart packages section along with $_EXTRA_RHEL_PACKAGES and $EXTRA_PACKAGES. The old packages list is in RHEL5.packages, with an initial list of RHEL6 in RHEL6.packages. The RHEL5 list has been slightly updated in the area of compat packages. The default value of $RHEL_PACKAGES invokes new function rhel_package_list(), which prints the contents of the packages file corresponding to the current $RHEL_VERSION. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Add RHEL6-compatible option to default value of KS_DONE_MESSAGE.Martin Schwenke2011-01-171-1/+1
| | | | | | | RHEL6 does not say "you may safely reboot your system" so we match "System halted." to know when to halt the kickstart guest. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Remove persisent NIC naming in basic-postinstall.sh.Martin Schwenke2011-01-171-0/+4
| | | | | | | | Different MAC addresses are used in guests for base creation than for actual clusters. We don't want the cluster nodes to remember any NIC/MAC associations from the kickstart run. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Add RHEL5.6 ISO details.Martin Schwenke2011-01-171-0/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Support for RHEL_VERSION=6.0 in rhel_print_iso.Martin Schwenke2011-01-131-1/+7
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Optimise newline subsitution during template expansion.Martin Schwenke2011-01-111-2/+3
| | | | | | | | | | | | | | | | | | The current code uses a bash variable pattern substitution. This can be horrendously slow on a large value. On my laptop it takes bash about 0.7s to do a single replacement of the 1st newline in a ~50KB value. Doing a template substitution during the generation of XML for 100 shared disks causes the creation of a 1000 line value that is about 35KB... so handling 100 shared disks can be very slow! :-( This changes newlines to character 0x01 using tr and then uses sed to replace these by "\\n". There are other subsitutions nearby in the code but they don't tend to cause this type of problem because they escape characters that tend to be used less frequently. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Allow a lot more shared disks (about 700).Martin Schwenke2011-01-111-2/+2
| | | | | | | | This is now a trivial change due to the use of arrays. Based on a patch from Eugen Anghel <eugen.anghel@ro.ibm.com>. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Change shared disk device name code to use a bash array.Martin Schwenke2011-01-111-4/+4
| | | | | | | Based on a patch for more shared disks from Eugen Anghel <eugen.anghel@ro.ibm.com>. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Revert "Allow a lot more shared disks (about 700)."Martin Schwenke2011-01-111-13/+2
| | | | | | This can be done much better using arrays. This reverts commit 5a5494821b51abfe2d6bf6f6dc9d960b8c3f0cde.
* Allow a lot more shared disks (about 700).Martin Schwenke2011-01-051-2/+13
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Added TODO file for cluster_configure.Martin Schwenke2010-12-061-0/+3
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Configuration file and wrapper script for simple GPFS-based NAS stack.Martin Schwenke2010-12-062-0/+161
| | | | | | | | Master configuration file for cluster_configure and a wrapper script to hold it all together. The script may eventually become part of cluster_configure. Signed-off-by: Martin Schwenke <martin@meltin.net>
* kickstart and post-install templated for simplified GPFS-based NAS cluster.Martin Schwenke2010-12-062-0/+333
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Initial version of cluster_configure.Martin Schwenke2010-12-0647-0/+1592
| | | | | | See base/all/root/scripts/cluster_configure/README. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Change the SCSI ID prefix in shared disk images from "SOFS" to "AUTO".Martin Schwenke2010-12-021-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* mknsd.sh creates its own list of disks to use as NSDs.Martin Schwenke2010-12-021-1/+3
| | | | | | | | | | | This currently uses /var/mmfs/etc/nsddevices to list the disks but this script isn't generally available. Instead we implement some custom logic that asks multipath about available devices and then filters on those created by KVM or via autocluster's iSCSI code. Signed-off-by: Martin Schwenke <martin@meltin.net>
* setup_gpfs.sh no longer uses cnpushfile.Martin Schwenke2010-11-291-1/+1
| | | | | | Use a more generic combination of onnode and rsync. Signed-off-by: Martin Schwenke <martin@meltin.net>
* New configuration variable GPFS_DEFAULT_MOUNTPOINT, used by setup_gpfs.sh.Martin Schwenke2010-11-292-1/+4
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Move GPFS config variables from 50sofs.defconf to new file 40gpfs.defconf.Martin Schwenke2010-11-292-6/+7
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Move public IP config variables from 50sofs.defconf to 00base.defconf.Martin Schwenke2010-11-292-6/+7
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Make 3rd octet of IP addresses configurable.Martin Schwenke2010-09-238-9/+23
| | | | | | | This introduces new variables IPNET0, IPNET1, IPNET2. These default to 0, 1, 2 respectively. That is, this change is backward compatible. Signed-off-by: Martin Schwenke <martin@meltin.net>
* NTP server should be GATEWAY, not a hand constructed IP address.Martin Schwenke2010-09-231-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* New -E options needs "shift 2" - oops!Martin Schwenke2010-08-251-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Add new -E option like -e but doesn't exit.Martin Schwenke2010-08-251-2/+5
| | | | | | | In complex configurations it can be useful to add a function in a configuration file and run it on the command-line. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix 2 bugs in iscsimultipath initscript.Martin Schwenke2010-08-121-1/+2
| | | | | | | | | * Incorrect reference to $c instead of $_c caused loop to never terminate. * Need to flush multipath devices before trying again. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix think in iscsimultipath initscript.Martin Schwenke2010-08-101-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Support HOSTS_STATIC_ENTRIES on the rhel_base nodes too.Martin Schwenke2010-08-061-0/+2
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* New configuration variable HOSTS_STATIC_ENTRIES.Martin Schwenke2010-08-062-0/+5
| | | | | | | This points to a file containing static entries that should be inserted into /etc/hosts on each node. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Rework the code that finds the list of variables in a template.Martin Schwenke2010-06-291-6/+7
| | | | | | | This version seems cleaner and clearer than the previous version, so is more likely to be correct. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix the template substitution in certain corner cases.Martin Schwenke2010-06-281-1/+2
| | | | | | | When template variables are surround by valid variable characters the substitutions can fail. This makes it more robust. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix hardcoded 10.0.0.1 in one of the iSCSI templates.Martin Schwenke2010-06-221-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* New function shared_disk_iscsi_setup_base_get_ip() prints IP for iSCSI client.Martin Schwenke2010-06-221-1/+7
| | | | | | | | | This make the generation of IP addresses for iSCSI client configurable. If, for example, you're using an IP addressing scheme that is out of autocluster's control, you can override this function to try and provide an alternate solution. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Register shared_disk_iscsi_setup_base in shared_disk_setup.Martin Schwenke2010-06-221-4/+1
| | | | | | | | | | | | | | Only register this is in setup_base_hooks if iSCSI shared disks are being used. It was registered globally. This may, this hook is registered even if you run clear_hooks on setup_base_hooks. This obeys the "principal of least surprise" in the case where iSCSI should work if you select iSCSI. However, it might surprise people to find a hook registered when they've clear the hook variable. Can't win 'em all. If necessary, shared_disk_iscsi_setup_base() can be overridden. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Move the "cluster created" message back into create_cluster().Martin Schwenke2010-06-181-1/+4
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Put /etc/hosts editing suggestion into cluster_created_hooks.Martin Schwenke2010-06-181-2/+7
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* In setup_gpfs.sh, getopt complains if it doesn't have a -o option.Martin Schwenke2010-06-181-1/+2
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Kickstart: don't install xulrunner-devel - nothing needs it.Martin Schwenke2010-06-181-0/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* grep sanity check in waitfor should now use extended regexps.Martin Schwenke2010-05-311-1/+1
| | | | | | i.e. Pass "-E" to grep. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Make waitfor look for a regular expression rather than just a string.Martin Schwenke2010-05-281-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Add $node_type to the UUID map.Martin Schwenke2010-05-281-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Modify create_node_COMMON() to pull the UUID for this node out of the map.Martin Schwenke2010-05-281-1/+3
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* Build a UUID map file in common_nodelist_hacking.Martin Schwenke2010-05-281-0/+8
| | | | | | | | | | | This maps between $ip_offset and UUID. This isn't used anywhere yet... however, there will be some autocluster applications where a node will need to know the UUIDs of other nodes. Note: after testing that the UUID map looks right, we'll actually use it in create_node_COMMON(). Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix a typo causing as syntax error.Martin Schwenke2010-05-241-2/+3
| | | | | | Also clean up some of the progress messages. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Support SYSTEM_DISK_MOUNTS in unmount_disk().Martin Schwenke2010-05-241-1/+12
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
* New configuration variable SYSTEM_DISK_MOUNTS.Martin Schwenke2010-05-242-11/+20
| | | | | | | | | Allows multiple filesystems to be mounted by mount_disk() at cluster creation time. This is useful if, for example, the base image has been created with separate partitions for / and /var, and you need to copy configuration to both filesystems. Signed-off-by: Martin Schwenke <martin@meltin.net>
* Fix typo "derconf".Martin Schwenke2010-05-211-1/+1
| | | | Signed-off-by: Martin Schwenke <martin@meltin.net>
generated by cgit (git 2.34.1) at 2025-08-31 05:07:26 +0000