Jac Backup Procedures

Published on June 2016 | Categories: Documents | Downloads: 21 | Comments: 0 | Views: 142
of 5
Download PDF   Embed   Report

Comments

Content

Joint Astronomy Centre Computer Services Group Backup Policies and Procedures

5 Sept 2003 JACSN/10.1

JAC Backup policy
The JAC Computing Services Group maintains backups of all computer data that are necessary to maintain the business continuity of the JAC. At least two copies of all raw data taken on JAC telescopes will be preserved for perpetuity. Backups will always include all users home directories, but will not include users personal data partitions or data on personal Windows-based computers. The procedures used to implement these backups accept the following limitations: 1. The speed with which data can be restored will vary depending on the length of time the data has been lost, but should be no greater than 24 hours. 2. Data that was created and lost within the same 24-hour period may not be able to be restored. 3. Backed up data that is not raw science data may only be recoverable for a period of one year after the backup date. The backup procedures also contain explicit measures to cover such unlikely events as catastrophic site loss, and malicious destruction of data, both by JAC staff and by external parties. The backup procedures, and the files that are backed up, will be formally reviewed every 6 months.

1

JAC Backup Procedures
1 Hilo Tape backup system
This system does a cycle of full backups on the order of every two weeks, and incremental backups every evening with the goals that data is protected against malicious or accidental corruption or deletion. The basic procedures are: 1. All data that is present on disk for at least 24 hours will be backed up to tape. 2. Any version of a file that was written to tape and is less than approximately two weeks old shall be recoverable from on-line tapes. 3. Once tapes are taken off-line they are stored in an off-site bank safe for a minimum of 1 year under control of the JAC Administration Group. 4. The list of backed up files is formally reviewed every 6 months. The files that are currently backed up are: • • • Every Hilo home directory. All ufs, ext2 or ext3 file systems on elua, hihi, malama and ola. Selected file systems on ieie, kiki, luu, nahoku, mauiola and ulili (See Table 1)

2 Hilo/Gemini disk backup system
This system (kaaohua) mirrors disks in the JAC Hilo site every 24 hours to a disk server at the Gemini base facility. The goals for this system are: 1. All data that is on the Hilo disk at the time of the backup will be present for at least 24 hours on the backup system. 2. We get rapid recovery of accidentally deleted files. 3. If the JAC Hilo site is lost then we have a current copy of Hilo disks for rebuilding systems. The basic procedures are: 1. The system uses a remote synchronization facility (rsync) configured so that an exact copy is made and files that were deleted in the last 24 hours will also be deleted in the copy. 2. The system is configured so a very limited number of people have physical an login access - all logins except root and members of the sysadmin group are disabled. 3. The backup disks are exported to Hilo machines read-only.

2

4. The list of backed up files is formally reviewed every 6 months. The files that are currently backed up are: • • Every Hilo home directory. All file systems on hihi, malama (except /budtool), and ola.

3 Scientific Data
All data is backed up to tape within 24 hours of it being taken. After this initial backup, at least two copies of the data are kept at all times – either two tape copies or one disk copy and one tape copy. Within 7 days these two copies will be kept in separate locations at least several miles apart to avoid simultaneous data loss. The definition of raw scientific data varies from instrument to instrument, and is defined at instrument delivery by the instrument scientist. Note: We are currently working on documenting these procedures more thoroughly.

4 Accounting System
A full backup of the system is done once per year to DLT (all partitions). At 10 pm every day we backup the /export/home/sun42/back directory to tape. This directory is updated when Velvet runs the daily macros. There are three backup tapes that are rotated each working day. Two of the tapes reside in the computer room. 1 tape resides in a bank safety deposit box. Velvet takes the current backup tape each morning to the bank and returns the other tape to the computer room. Every day we also backup the entire /export/home/sun42 directory to the backup accounting server 192.168.90.11 (kualua). This machine is a hot standby.

5 Timesheet system
The backup the SQL web timesheet database The SQL server runs a full backup at 10 pm Monday to Friday, and then a scheduled batch program copies the backup to ola with a distinct timestamp at 11:45 PM. The files on ola are purged periodically, but all the files backed up on Fridays are retained, and they also are backed up to tape as part of the tape backups.

6 OMP Sybase Data
The complete database is replicated from UKIRT to JCMT to protect against network, equipment or site loss. The database is backed up to the /export/home disk partition on mauiola daily and this partition is also backed up to the Hilo tape system.

3

7 Archive Sybase Data
Both the JCMT and UKIRT archive databases are mirrored to external sites daily. The database is backed up to disk daily and these disk files are also backed up to the Hilo tape system.

8 CVS Repository
In many ways the files in the JAC CVS repository are the “Crown Jewels” of our data, in that they are the data set that we have expended the most effort in generating, and would have the greatest difficulty replacing from scratch. Whilst they are backed up in the normal process, we have also taken steps to protect against malicious damage: 1. Different people take multiple backups. 2. At least one of these backups operates on a two hour cycle to an off-site location (ATC, Edinburgh). In addition, multiple copies of the files in the repository are checked out at various sites around the world.

9 VAX systems
Summit VAX system and software disks are backed up to hot spare disks weekly. Tape backups are made occasionally. Note that the VAX backups are currently being reviewed. Possible changes being considered include: 1. Replacing summit VAX file-servers by Linux VAX emulators, and backing up these Linux machines to Hilo tape sub-system. 2. More regular tape backups 3. Some other form of off-site backup – putting a VMS backup save set on a backed up Unix partition regularly?

10 Disk mirrors
In addition to normal backups, much of our data is protected against data loss by RAID5 (for bulk data) and RAID 1/2 (disk mirrors – for operating system file servers).

4

Appendix 1. List of backed-up file systems
Table 1: List of partitions backed up on JAC servers apart from servers that are fully backed up Host name ieie Partitions backed up / /export/home /export/Linux/star /export/Linux/local /jac_sw /jcmt_sw /local /local/WWW /observe /star / /export /export/home /export/observers /opt /local /star /ukirt_sw /jac_sw /export/Linux/local /export/Linux/star / Partitions not backed up /opt /epics_data /jcmtarchive /export/radiometer

kiki

/export/ukirtdata /export/epics_data /export/WWW

luu

mauiola nahoku

/ /WWW /export/home /var/spool/news

ulili

/ /WWW

/scubadata /scratch /export/DSS /export/tempdata /export/ukirtdata /export/jcmtdata /export/tmpdata /smartcd /tmp /DSS /cds / /export/home /export/data /nt /tmp /export/home /nt4 /tmp

5

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close