Backup of a VM to local and remote repository

[PerO]

Hi,
We have a setup where we have two offices, on small and one larger (hub).
On the small office we have some VMs running on a single host and one backup server with large disk.
On the hub we have a cluster of VMware hosts and one backup server with a lot of disk and a tapelibrary.
The wan-link is only 4 Mbit/s
On both backup servers are we running VeeAm 7 (std) and my intention was to make a local backup of the important servers and also a backup to the hub office.

I have set up two jobs, one to each location and I did a preseed to the hub and everything looks ok, until I checked the backup sizes.

The local backup is running first and have some Gb. The backup to the hub-site is much smaller and I "guess" it is because of CBT.
The local backup is "reversed incremental" and the other to the hub-site is "incremental forever" with synthetic full on Saturdays.

How shall I set it up so I have useful backups?

best regards
Per

[foggy]

Per, what is your concern here? CBT should work fine if you run multiple jobs against the same VMs. Could you please clarify what files do you compare? Is it the size of the entire directory or just the size of the latest restore point?

How shall I set it up so I have useful backups?

Btw, if the job has Success status, then you do have restorable backups.

[PerO]

Hi,
The first job, running at 20:00 with no one in the office shows 1,8 GB transferred to disk

Name Status Start time End time Size Read Transferred Duration
TH1-FS001 Success 20:00:40 20:09:54 580.0 GB 4.4 GB 1.8 GB 0:09:13

The second job, running at 22:00 with still no one in office shows: 301 Mb transferred to remote site

Name Status Start time End time Size Read Transferred Duration
TH1-FS001 Success 22:01:04 22:29:22 580.0 GB 939.8 MB 301.3 MB 0:28:18

How can the second backup of the same server show that much smaller amount of data read and transferred???

[foggy]

Are all other settings the same for those jobs (e.g. CBT, swap file blocks exclusion, probably some disks excluded in the second job)? Could you give both jobs another try and see whether you see the same difference in the data read?

[PerO]

Hi,
Yes, I think so. At least my intention was that.
Local is as I wrote Reversed Incr and the remote is forewer incremental with synt full over weekend.

I have changed the order and perform the remote backup first and then the local. The server is located in Thailand and Malaysia and their working day are starting soon. Lets see tomorrow night.
thanks Per

[PerO]

Now have I changed the order and do the backup to remote site first and some hours later (late evening local time) the local backup.

3 กันยายน 2013 19:00:09
Success 1 Start time 19:00:09 Total size 580.0 GB Backup size 345.9 MB
Warning 0 End time 19:30:57 Data read 957.3 MB Dedupe 1.0x
Error 0 Duration 0:30:47 Transferred 329.0 MB Compression 2.7x


3 กันยายน 2013 23:45:22
Success 1 Start time 23:45:22 Total size 580.0 GB Backup size 95.1 GB
Warning 0 End time 23:51:29 Data read 1.5 GB Dedupe 4.8x
Error 0 Duration 0:06:06 Transferred 615.3 MB Compression 1.3x

The strange thing is that the size of Data Read is almost 2 times larger when I do the local backup. 957 Mb vs 1,5 Gb
Why is it so big difference in Dedup and compression?

Anyone know how to change the date format. From the beginning the server has Thai regional settings but we have changed that to English but months are displayed in Thai (i guess )

[Vitaliy S.]

Dedupe and compression ratio is calculated per individual restore point. Since you're using two different backup modes - forward incremental and reversed incremental, the dedupe and compression ratios are displayed for VIB (incremental point) and VBK (full backup) and the difference is absolutely expected in this situation.

The strange thing is that the size of Data Read is almost 2 times larger when I do the local backup

Why is it strange? As far as I understand your local backup now starts later, so it should read more changes that occurred since the latest backup, right?

I have set up two jobs, one to each location and I did a preseed to the hub and everything looks ok, until I checked the backup sizes.

Again, the difference in the counters (read/transferred) can be explained by the different state of the initial full backup. What backup files did you use to pre-seed your job targeted to the offsite location?

[PerO]

Dedupe and compression ratio is calculated per individual restore point. Since you're using two different backup modes - forward incremental and reversed incremental, the dedupe and compression ratios are displayed for VIB (incremental point) and VBK (full backup) and the difference is absolutely expected in this situation.

Ok, got that.

The strange thing is that the size of Data Read is almost 2 times larger when I do the local backup

Why is it strange? As far as I understand your local backup now starts later, so it should read more changes that occurred since the latest backup, right?

Ok, make sence but there is no activity on the server at all at that time. And why was it the other way when I have the jobs in the other order?
Local backup at 20:00 was reading 4.4 GB and the second backup to remote repository was only reading 940Mb[/quote]

I have set up two jobs, one to each location and I did a preseed to the hub and everything looks ok, until I checked the backup sizes.

Again, the difference in the counters (read/transferred) can be explained by the different state of the initial full backup. What backup files did you use to pre-seed your job targeted to the offsite location?

I created a second repository on local server and made the first backup (incremental) to that. Copied the files to external disk and sent it to Malaysia. Changed the location of the second repository to the server in Malaysia and then copied the files from external disk to this new repository.
After some days of backups turned on Synthetic Full on Saturdays. The only difference between the jobs is that I have optimized for WAN on the remote backup (and compression rate, Optimal vs High)

[Vitaliy S.]

Ok, make sence but there is no activity on the server at all at that time. And why was it the other way when I have the jobs in the other order?

Are you sure you do not have any maintenance tasks at this very time? Please take a look at the existing topic for more information on this > Is it possible to see/analyse contents of a backup file?

After some days of backups turned on Synthetic Full on Saturdays. The only difference between the jobs is that I have optimized for WAN on the remote backup (and compression rate, Optimal vs High)

How many job runs did you already have? If you continue running these jobs, then you should see approximately the same changes (read) for both jobs, the only difference in the amount of sent data can be caused by the difference of the job start times.

[PerO]

I have now run both jobs again and with only 1/2 hour inbetween.
First job to local rep and that read 1,0 Gb
Second job to remote site read 598 Mb

I have run this remote job since 15 of August (total of 11 times)

[tsightler]

This sure looks like the difference between WAN and Local block size settings. Are you 100% sure that both jobs are using the exact same settings for Storage Optimizations?

[PerO]

Did a quick rerun and now does the local backup read 270Mb and the remote only 52Mb.

There is different settings in the jobs, Remote job optimized for WAN and has High Compression
The local backup is optimized for Local Target and optimal compression

but that shouldnt effect the READ size, or?

[tsightler]

Yes, that's exactly what this setting affects. The Local/LAN/WAN settings actually change the block size that Veeam uses when reading changed data from the source, and correspond to 1MB/512KB/256KB sized blocks. If a single byte is changed on disk, Local setting will read the 1MB block that contains this change, while WAN setting would read a much smaller 256KB block, so WAN will read and process a lot less data, but at the cost of higher CPU, memory usage, and I/O operations on the target, which is generally a good trade for WAN. In practice the difference between Local and WAN won't normally be 4x because odds will say that on disk changes will be spread across 256K blocks that are "near" each other and thus would be part of the same 1MB block, however, based on I/O patterns and change rate it's not uncommon for the difference between the two settings to cause a 2x reduction in the amount of data read.

[PerO]

Ok, got it.
thank you for the info.