There is work to be done! There's a war to be won!

Thursday, 2 January 2014

Symantec NetBackp 7.6 - Sexy integration continues with shiny new features.

Happy New Year to all ! Truly hope that its filled with everything you want for yourself; importantly peaceful, healthy, kindness ... all the good things. To kick off the NY, I thought I'd dump my 2cents worth on Symantec new NetBackup 7.6 offering. FA (first availability) was in Oct '13, with GA announced in December ( Goodness me, what a lot of exciting features await.

To begin with, MSDP, whether integrated as part of the NetBackup Appliances or not, is certainly sounding a lot healthier with 'self-healing data integrity checks'. Too good to be true? Not from what I've seen: As part of a code rewrite, MSDP-DB (dedupe db) has moved from the 7.5 segment based offering (recall the dedup plugin divided the file into segments and sent those unique segments to dedupe engine (spoold)) - we now use a multithreaded agent (inline) btwn dedup plugin and dedup engine using a referencedb (ok, its a flat-file but anyway), it replaces postgresql and is container-based. So with spoold now containing refdb 'files' we're no-longer concerned about transaction logs in postgres - none of this post-processing stuff that slowed everything down. We're also limiting any corruption to a container of refdb, which is going to be a lot better than the previous single point of failure in 7.5 where nothing started if you were corrupt. What a pleasure. This is also why I saw a huge increase in start-up times of spoold under my little DAS running MSDP - it was faster, no question about it.

* So: things to look out for:

mtstrmd (multithreaded agent)
dedupe-engine(spoold) now has a 'refdb' (dbpath/refdb) with sub-directories that contain the id's (refid.ref).

SLP windows. Yes, I hear you cry:

- Scheduled windows can now be applied to dedup operations and/or replication operations (AIR, not OST). You can also delay the duplication of an image until source copy will expire (4 hours default)
- The GUI has finally got SLP parameters ... forget the config file, click your way into '14 ! lol
- I couldn't test the logic around processing to a great degree as my poor DAS was in a state of shock :) 
Jobs were doing what they were told to do... I kicked off an many as I could handle.
- I did find some welcoming activity monitor progress and size stuff, lovely.
- Can auto-resume files at a selected time or suspend. (however: cant suspend msdp, ndmp, opt dup, AIR (!) or backups from snaps ... lol) - read that manual.

OpsCenter - its all heading here, so get going:

- Security enhanced w/integration AD/LDAP (ad groups) and better user role assigment.
I also read with interest about the embedded authentication broker: each ops server has its own local server, so simplifies issues around install/config/upgrade/uninstall. Also other prods, if any, are not affected by removal, we hope.
- Cool getting started wizard. I always think these things need to exist to make the product more accessible
- Fully NBU managed (no further support for Tivoli, CommVault, EMC Networker) - fair enough.
- Telemetry.
- 'NetBackup Appliance aware' - not tested as Santa didn't buy me an appliance. Its said to have some enhanced monitoring and alerting regarding h/w failures - certainly has a new monitor appliance h/w tab. I saw a large number of db tables associated with Appliances, so I'd bet its all tightly taking care of any issues and sending on the alerts. (cpu, fan, disk, hba ... pci ... raid .. powersupplies ... the list goes on ... fill your boots)

* New daemons & stuff:
opsatd (embedded AT broker)
Logging still the same, but appliance logs will probably be of benefit to those running them by searching for a keyword associated with refs from the db table (all appear to start with am_HardwareFailureAlert, am_CPUFailureAlert, am_DiskFailureAlert, etc)

Oracle Intelligent Policies:

My testing was conclusive on this one - it allowed the automagical discovery of the oracle instance I was running. I backed up the DB, control file, logs, once I had completed registration; this registration requires you to provide credentials that are stored in NBDB. These credentials can be OS login or Oracle login - this is where you can define an instance group to hold the container of instances and apply it elsewhere later. You can use 'auto-registration' for 100's of other instances if you need to. Lots of stuff to consider here - read the manual, slowly !

A new auto schedule 'archive log' is now available with a minutes frequency: re-do logs are backed-up :) Loving that.

* New daemons and more stuff:
- GUI driven new policy type is Oracle. Well laid out and easy to get going.
- nbdisco for instance discovery and mgmt (master/client)
- use regular nbpem, nbjm, nbsl, nbars for job discovery and admin log, also bphdb
- nboraadm : new cli to play with instances, else just use GUI.
- bpplinfo has some additional switchs, but we're told to use them with caution. -client_list_type and -selection_list_type (0 1 2 3 4) (legacy, wholedb, tablespace, datafile, fra) - nice.


Its taking awhile, but great to have read about wildcard, all_filesystems and exclude lists now being available for NAS. Note: NetApp works with a volume and directory, all other vendors only currently support volume interrogation.

The ALL_FILESYSTEMS is like the ALL_LOCAL_DRIVES, but the VOLUME_EXCLUDE_LIST must precede the ALL_FILEYSTEMS directive:

e.g. VOLUME_EXCLUDE_LIST = /vol/somedir1*, /vol/somedir/[2-5]*, etc


This came out in 7.5 and allows you to discover data/index it, and put a hold on it for legal reasons, etc.
The problem in 7.5 was that the index server had to be a media server. The ability to install this as a client (Windows 2008/2012) certainly helps with the one less license, even though the index server is a licensed feature. The main process here is searchexecutor.exe (aka: velocity). I originally tested this on 7.5 and it worked with a little bit of tweaking, but didn't get to test it on 7.6

In OpsCenter, there is search&hold tab that allows you to search for images based on data range and puts a hold on all images. Useful stuff. You can also integrate it with EV. I get the license thing but I wonder how long the license will be a requirement. Cant help thinking about competitors ... is it time to make things even more attractive by offering data discovery and indexing as part of the core product? I think so.

VmWare w/Accelerator & Instant recovery

Accelerator was introduced in 7.5. It tries to increase the speed of backups by using a tracking log on a client, and then sending those changes/logs to the media server. It will also use the windows change journal if its enabled (or, um, not enabled). Regardless, the idea is 'faster' backups and useful for clients with a low data growth rate. So do your research before turning it on - know your data, test it in your environment and hopefully you'll be synthesising full backups in no time at all. Also, have a traditional full backup of this client running once a week anyway if you do choose to use it ;)
So, with the power of our VmWare/VSphere friends and Vmotion coupled with the ever-cool-titled 'instant recovery', we simply turn things on: the disk storage on the appliance or the media server is handed over to the ESX server's datastore and the VM backup-image is 'recovered' (powered-on). Its all NFS here, so do your homework. A plugin also exists for vCenter allowing backup history lookups, monitoring and recovery. A few clicks and your favourite VM admin can be viewing the wonders of a window that looks similar to the BAR.

Windows 2008/2012 and Solaris 11

Solaris 11 (root) ZFS changes have been given the all clear with Sun Cluster - I do wonder how many people are still running their core servers on Solaris. Windows 2012 and 2008 support too with NTFS dedup (OS dedup) supported - makes me think how many people will use this on the OS or just allow the storage to handle it ... certainly will cause some issues with backup tools, where arguably OS dedup schedules will run with backups at the same time. Headaches.

New catalog compression

I did notice a different compression being used on NBU catalog - wasn't able to find out much more about it, but was told it offered better access times and told it favours 'lots of small images'. I also read up on the fact that EMM and NBDB have to reside on the same server going forward. Never had a need to install EMM anywhere else myself.

There is also good mention about the additional pre-checks done in this version, and I must admit that it really is about time that things like EEB's are checked and validated.

Well, hope some of that has been interesting to someone out there.
As always, I tend to do all this for myself as I'm genuinely interested and like many people tend to forget things ... which is what most of this blog is all about at the end of the day. Drop a comment if the urge, um, urges ;)

Quick links:
7.6 Docs -
(all 7.6 in 1 file)

Happy data protecting in '14 and best of the best as always.

No comments:

Post a Comment