Parallelism exceeded, during SQL backups

Hrvoje, we use a capacity licensing model, where we are charged solely on the amount of data we protect.

I install the Networker Agent and SQL module onto each SQL server. Then, I add that client to the SQL backup protection group, and point the saveset to MSSQL: Then, I have separate policies for the tlog jobs, grouped by their daily frequency. The backups are going to DataDomain. Can I just go into nsradmin and increase parallelism to 128 or whatever value I need until the errors stop happening?

Thanks.

Are you doing instance level backup or you name DBs for DB and TLOG jobs?  If you use DDBoost, you can still keep single server as streams go directly to DD, but you must add storage node license to bump up stream counts.  This means if you use legacy license model, then your server edition is limited to 64 streams, but with each storage node license you can bump it for additional 32 streams (regardless if you have or have not storage node).  With capacity license, you can go up to 1024 streams so if you suspect streams on NW to be limit, by just tweaking licenses you can fix it.

Ah, perhaps one more thing.  Once upon the time I noticed that my server (and my storage nodes while I used them) used parallelism of client (client for backup server and/or storage node).  Since then, if I have only backup server, I set client parallelism of backup server client to the same value as noted in server parallelism.

Yes you can.  You should be able to increase it without issues.

Thanks Hrvoje. I changed the parallelism setting in the gui to 1200 (just to see what my maximum was), and it said I was outside of the range of 1 to 1,024, so I can see that my max is 1,024. I changed it to 128, and will start from there. I do see that each client under client settings, has its own parallelism setting under the Globals 1 of 2 tab. They all seem to be set to 4. Is this the other parallelism setting that you are referring to?

Regarding server parallelism, this was (historically - I'll say this was the case in 2005 for sure, but I *doubt* it's changed) set against the number of incoming *NetWorker* backups that it would accept at one time (save processes or similar). 

Application backups (such as SQL Server, Oracle, Exchange, or whatever), while launched from NetWorker, are fully controlled by the internals of how that application runs its backups.  NetWorker may schedule and control their launch and pass them some backup parameters, but the backup itself is application-native and falls outside of NetWorker's server parallelism limitation.

Toddman214 wrote: Thanks Hrvoje. I changed the parallelism setting in the gui to 1200 (just to see what my maximum was), and it said I was outside of the range of 1 to 1,024, so I can see that my max is 1,024. I changed it to 128, and will start from there. I do see that each client under client settings, has its own parallelism setting under the Globals 1 of 2 tab. They all seem to be set to 4. Is this the other parallelism setting that you are referring to?

Yes.  I usually for DB client set this anywhere between 16 to 36 - depending on what is the client and how many databases or parallel streams I have. Note that you should also not push things to the limit as your DD might have session limits as well (this entirely depends on DD HW and sometimes on DDOS too). For backup server (and/or storage node) client, this value should be high as my previous experience shows that storage node was limited in writes or let's say it was overriding server parallelism (that in my view is bug, but since I now mostly use single backup servers as vm I set this for backup server client to the same value as server parallelism and if I need to control number streams I set this normally on group level).

Hi Todd,

          Best bet is to keep the Server parallelism to maximum (which is 1024). Then modify each of the client parallelism according to your need. If you have about 100 DB's in the SQL client then you can increase the client parallelism to 100.

Hi,

You have an interesting problem. My problem was how to limit the nr. of sessions of MS backups because NMM module doesn't honours client parallelism settings and our servers were a bit overloaded by the backups.

(I don't wanted to create many savegroups with a limited nr. of DBs etc., I like if I can set "backup all DBs", because this way when a new DB is created then there is no need to worry about the backup - it will be automatically included / and no failures on DB removal).

The resolution was to create a new pool and 1 new dd volume with a lower number of max sessions, so networker doesn't complains about anything, he simply queues the sessions.