I have been building a new NetBackup implementation to backup an NetApp Filer which uses a 5230 appliance as the Master Server and two large Windows based Media Servers. The appliance was shipped in Dec 2015 and I have upgraded it to v2.7.1 specifically to take advantage of the Accelerator for NDMP which is available in the bundled NetBackup v7.7.1 version.
The primary objective is to create a backup scheme which is tapeless. The source data is about 6TB on a single NetApp Filer, the Netbackup 5230 appliance has 13TB nominal usable capacity and the two Windows based media servers have 70TB usable each. The design has attempted to provide some resilience against data loss in the event of a major building level disaster by locating the Media Servers in different buildings as a short term poor man's Disaster Tolerance until next year's budget can purchase the second appliance and a more complete solution provided. There are a small number of conventional Windows agents too, but the main focus of this post is the NDMP backups.
With only one Master server, I cannot use AIR and intended to use the appliance as a Master/Media server and use Storage Lifecycle Policies to copy the backup images onto the Media Servers. I have learned (the hard way) that I cannot use the approach which I originally intended and use the Master/Media server on the appliance as a "staging" area and use a SLP to copy the backup images to the large media servers and "Expire after copy" on the Appliance because the Accelerator needs access to the previous backup image to be able to work. This was diagnosed from the error messages where the BASE_DATE could not be determined. The backups must go directly to one of the two Media Servers and an SLP copy the image (for a poor man's Disaster Tolerance) to the other media server. The Appliance will now play little part in the solution design, except to act as the Master Server.
I am concerned about the track logs which are used by the Accelerator.
I have discovered the track logs "track_journal.v2.dat" which are stored on the Master and Media servers by reverse engineering the instructions for relocating them elsewhere. So far I have not been able to find anything authoritative which will allow an estimate of their size, but by experiment I am beginning to get a feeling for them.
I am now aware that there are no only a copy of the track log on the Master and Media server but that there needs to be room for the file from the previous backup but also another file of the same size which is built by the current backup. On a Master/Media server there are copies of these files in netbackup/db/track and netbackup/track.
Where the Master/Media server is an appliance, I can find no GUI tools available which monitor this use of disk space. In particular the file protection scheme only allows an account which uses the NetBackupCLI method to even list the contents of the netbackup/db directory.
I can see that following experimental backup policies and schedules there are multiple files in the track log locations on the appliance and media servers which are no longer required. My primary question is how to housekeep these files?
1. Will track_journal files get automatically deleted by the NetBackup system when the corresponding backup images have expired and there are no longer any entries in the Catalog?
2. If the track_journal files will not automatically get deleted, can these be removed manually?
3. Where the Master server is an Appliance, what is a supported procedure to remove these files manually as they are owned by root and protected against deletion?
I appreciate that the Accelerator for NDMP is a new feature and that Symantec were previously rather secretive about the internal workings of a proprietary technology, but the lack of information about it makes delivering and supporting solutions which are based on it particularly difficult. I would appreciate some fairly authoritative comments from the Veritas development and support teams to throw some light onto these issues.
Best regards - Steve Dawes
