Understanding how the Backup Service Works

There are lots of posts that describe how to configure and setup pool pairing including my own description.  But what I haven’t seen is a good description of how the backup service works or what is happening behind the scenes.  So in this post I’ll attempt to walk through the process, what is happening and some of the under the hood that I could discover through investigation.

The Backup File Share

The heart of the backup process is the file share.  This will be located in the same location as your normal Lync file share.  So in my case, I have two standard edition servers in a pool-pairing relationship.  Here we can see the contents of that folder:

Untitled

Here we see four folders.

CentralMgmt is going to contain the files necessary for the backup and replication of the Central Management Store.  Two items of note here.  First, if the servers in your pool pairing relationship are not the CMS server you will not see this folder.  Second, replication of CMS for the purposes of backup is completely different than CMS replication that occurs throughout Lync.  The Master Replica and Replica Agents are responsible for keeping the RTCLocalxds database up to date.  The backup service is responsible for keeping the RTCxds database up to date.  Presumably these two databases are nearly the same in contents (the RTCxds since it can take over as CMS for the organization will have more details about replica’s, etc.) but they server very different purposes.

ConfServices are for the replication of conference data from the file share.  When an individual uploads an attachment into the conference we want to make sure that same data is available on the far side in the event of a pool failover.  This is in particular helpful for those individuals who pre-load content.

UserServices are for the replication of buddy lists, conference directory ID and other user services information.

You will also find a TaskAssignment.  I cannot find any purpose of the TaskAssignment folder as it’s contents never seem to change.  What is located in that folder is a text file with the servers name listed.  Occasionally a Temp folder will also appear and it is exactly what you would expect it to be used for.

Inside of each of these folders is a Cookie and Data folder.  The Cookie appears to keep track of when the last time data was sent and received.  The data folder contains the actual data being replicated.

File Copy Process

Similar to the CMS replication process, the actual process of copying data from PoolA to PoolB is via an SMB file transfer.  What is different however, is that it appears the backup service itself is responsible for both the file copy and consumption of data.

Typically the folder structure will be empty in the data folder but in an attempt to capture the data let’s do some experiments.  In my lab the servers are lyncfe03 and lyncfe04.  So we will stop the backup service on lyncfe04 and invoke a backup sync.

Some observations:

1. After the invoke, the LyncFE03 server copies the file from FE03 to FE04.  Since the backup service is stopped the file share isn’t being actively checked for files/changes so they remain in place.  But you can see that LyncFE03 needs permissions to write directly into the LyncFE04 file share.

pic2

2. The PresenceFocusCookie folder also has a cookie.zip file as well.  Exacting those files reveals a series of folders and XML files for consumption into the backup service.  Based on viewing the contents of the XML file it doesn’t appear anything is written in “plain text” and somehow the data appears obfuscated to some degree.

pic4

Status of Backup

Based on the investigation of the backup service it’s critical that when you use the get-csbackupservicestatus command that you run it on both pools involved in the pair.  For example, with my backup services stopped on lyncfe04 still, I get the following output.

pic3

We can see the first command is the successful output of the status of lyncfe03.  Here we can see the export status (export to Lync FE04) is in a steady state.  And the import status (import of Lync FE04 data) is normal.  What is doesn’t tell me is that the LyncFE04 server has actually done anything with that information.  The second command which bleed red all over the screen is the fact that the backup service is down.  Once I start the backup service again on LyncFE04 and allow it to watch/consume data again the backup.zip file disappears.

Back on LyncFE03 however, the PresenceFocusCookie folder was empty when I did the invoke.  Once the services started again, the cookie file is back.  So it appears as though the servers write a file back.

Configuration of Backup

One item to make sure to pay some attention to is the actual configuration of your backup service.  You may want to refine this depending on what your WAN bandwidth that is available between location.  A get-csbackupserviceconfiguration reveals:

pic5

Here we can see that by default, your backup service will sync information every two minutes.  Additionally, it has a maximum conference size of 100Mb.  This could be important if your end users have a habit of uploading large files to your meetings and you want to ensure you don’t crush your WAN.

Conclusion

So what did we learn within this testing.

  • The backup service runs as a service on front-end pools when pool pairing is enabled.
  • This service is both actively copying data and consuming data
  • The backup service appears to track changes to the blob store in the SQL database and writes that information into files on the far side server
  • The backup service is going to monitor its own file share and write that information back into its own database.  Presumably, if you had an Enterprise Edition pool, the windows fabric process would update the remaining servers (I’ll do some more investigation to verify).
  • Remember, that you can configure all of the backup settings if you want.  So outside of the initial sync, the process isn’t overly heavy on the WAN (with maybe the exception of what could be conferencing data).

Share

0 comments on “Understanding how the Backup Service Works”

  1. Pingback: Understanding how Windows Fabric Works

  2. Pingback: Understanding how Windows Fabric Works (at least in regards to what Lync uses it for)

  3. Chandru Reply

    as per article the “backupstore” folder stored in File store location.

    are these data belongs to A site or B site?

    if these data belongs to A site, where is B site data?

    in case of B site down, from where A site FE servers access B site data?

    • Richard Brynteson Reply

      The data is going to be stored in each pools file share. So the backup service on Pool A is going to copy (export) the data into the file share location of Pool B. Pool B will do the same back to Pool A’s file share.

  4. Chandru Reply

    As per my understanding the master replica agent running on the node who owns the CMS database in SQL server downloads a copy of SQL database and replicate itself (RTCLocalXDS) to all other Lync servers.
    the Replica Agent service running on all Lync servers will writes those information to its own SQL database RTCLocalXDS

    Please correct me if am wrong.

    also

    You mentioned “”The backup service is responsible for keeping the RTCxds database up to date.”””, In case I have one Lync site with one FE pool (No pool pair and No backup service), that time, how the RTCXDS database in SQL backend server gets updated???

Leave a Reply

Your email address will not be published. Required fields are marked *