Linux Server Configuration
This page will be updated with configuration notes during the process of configuring the first new clone of viking.
Raid Configuration Notes
Raid array configuration is managed in a bios-like configuration system entered during the boot process. There is a notice to press (Ctrl+R) to enter the configuration utility. On Viking Ctrl+H accomplishes the same goal.In order to create a new array a new configuration is made. By default the installed disks show up as JBOD. They must be made available, where the JBOD disks are listed there is a "Make good" action which frees a JBOD drive to be added to a new array.
After creating the configuration a new device will show up in the OS. There does not appear to be any control over what letter device is created (eg /dev/sda vs /dev/sdb). I accomplished partitioning and mounting the array as follows:
Be careful about this!! Ensure that it is done on the correct device (/dev/sda sdb, etc)
umount /mnt/data1 parted /dev/sda ....rm 1 ....mklabel gpt ....unit TB ....mkpart primary 0.00TB 64.0TB ....quit mkfs.ext4 -L raid-array-1 /dev/sda1 mount /dev/sda1 /mnt/data1
This will create a new partition labeled as "raid-array-1"
Updated /etc/fstab as follows (per notes, need to check)
LABEL=raid-array-1 /mnt/data1 ext4 noatime,defaults 1 1
Hot Swapping Drives
Drives can be added/removed while the system is online. I have not tested this with raid array members. Inserting a SATA drive into one of the available bays causes the drive to appear as a new device. The new devices must be manually mounted. To avoid conflict with the boot drive or other drives this should be done while the system is online and any permanent mounts should use the UUID or device id as visible under /dev/disk/by-uuid/Data Sync Across servers
- Globus is faster than rsync. Currently still testing, will update this section when globus testing is done* This command will synchronize a directory from viking to valkyr. The options of rsync permit merging multiple directory structures into one. Options also include resuming partial files and checksum validation.
rsync -rgptuP --ignore-existing --append-verify /mnt/data2/3SDW/3SAQS/2011 dschmidt@valkyr.cira.colostate.edu:/mnt/data1/WAQS/ du /mnt/data1/3SDW/3SAQS/2011 -lhP --max-depth 0
File System Permissions and Ownership
Warehouse data partitions are owned by root:3sdw on viking and root:iwdw on valkyrThe following permissions should be maintained on the main data storage partitions. ug+rwx o+rx o-w
Occasionally when new data is loaded it is necessary to reset the permissions. as follows:
chmod -R ugo+rx /mnt/data1
failure to include the read option for other will prohibit users from downloading the data. Failure to include the execute option for other will allow users to download the data but will give them permissions errors when they attempt to access downloaded files. If this occurs it is sufficient for the user to execute the above chmod command on their own copy of the downloaded data.
Globus Endpoint Configuration
https://docs.globus.org/resource-provider-guide/#install_section https://www.globusid.org/loginNeeded to install epel release
curl -LOs http://toolkit.globus.org/ftppub/globus-connect-server/globus-connect-server-repo-latest.noarch.rpm rpm --import http://toolkit.globus.org/ftppub/globus-connect-server/RPM-GPG-KEY-Globus yum -y install globus-connect-server-repo-latest.noarch.rpm yum -y install yum-plugin-priorities yum -y install epel-release yum -y install globus-connect-server
Next configure /etc/globus-connect-server.conf with globus account credentials, default folder, folder restrictions.
Run
globus-connect-server-setup
This will cause an endpoint to be listed in the Transfer files web interface. I made the error of attempting to add a new endpoint but was never able to activate it. Need to use the autogenerated endpoing iwdwcira#[name].
Endpoint is activated with ssh credentials. This could probably be done with ssh keys instead to prevent the need to activate. Using activation there is an expiration time.