Unsolved
1 Message
0
638
May 26th, 2022 04:00
Networker 19 + DD cloning sessions
Hi All, relatively new to Networker and DD, I'm expecting to have missed something obvious here.
When running a clone of ~250 savesets from a single device on an on prem DD to another single device on a DDVE in AWS, I seem to hit a limit of 30 ddboost repl-out/in streams active between the two datadomains. In the clone details Networker shows a single Clone session, with 100's of save sets "waiting to run" and 29 save sets "currently running".
Networker parallelism is set to 1024, and I've tried every combination of target/max sessions on the source and target devices, none of which seem to move the needle.
Any advice for how I increase this?
No Events found!
bingo.1
2.4K Posts
0
May 26th, 2022 05:00
The number 1024 is related to the total of simultaneous streams in the complete data zone. They are usually spread over multiple devices. If you have only one device, you will most likely end up with your scenario.
So the best is to add more devices so taht multiple clone jobs can run at one time.
barry_beckers
393 Posts
0
June 6th, 2022 08:00
you might want to look into the NW nsrcloneconfig debug option as stated in the NW admin guide, specifically into the values picked up by having concurrent_clone_enabled (default is YES) and how it relates to other values that can be set in /nsr/debug/nsrcloneconfig.
I also assume the clone action is set high enough? Set to the default 100? If I recall correctly 30 is the default amount of clone sessions, this not to cause too many problems on the DD's, running out of sessions.
"concurrent_clone_enabled
This flag need not be set in the nsrcloneconfig
file. The default value for this flag is TRUE.
It can be changed to FALSE by setting
max_threads_per_client and max_client_threads explicitly
to 0 in nsrcloneconfig file. When concurrent_clone_enabled
is TRUE and max_threads_per_client, max_client_threads
and max_savesets_per_thread is set to non-zero values,
then the clone parallelism is computed such that
max_client_threads *
max_threads_per_client <=
max_concurrency
max_concurrency =
MIN(30, dest_pool_parallelism,
available_stream_count)
It is recommended to modify RPS disabled DD to DD proxy
cloning."
"The nsrcloneconfig file enables you to add debug flags, control cloning sessions, and use the AMS functionality. It must be manually created under the /nsr/debug folder."
AMS (automated multi-streaming) might not be advisable to user towards GCP, as it would require 10Gbps connectivity between both DD's besides the other requirements below:
"AMS is supported only if:
● Both the source and destination Data Domain systems support the virtual synthetic capability (DDOS 5.5 and later). This can
be validated through ddboost option show command as shown below:
ddboost@localhost# ddboost option show
Option Value
---------------------- -------
distributed-segment-processing enabled
virtual-synthetics enabled
fc enabled
global-authentication-mode none
global-encryption-strength none
● The save set file being copied is large enough for the use of AMS to provide an improvement over normal replication.
● All save set types other than VBA, vProxy, and synthetic full. The exception is for Microsoft NMM Exchange module save
sets, where AMS is used even though it uses BBB and synthetic full.
Enable AMS, if the underlying bandwidth between two DDRs is 10Gbps. Because the use of AMS creates multiple streams, there must be enough bandwidth between the two DDRs being used for the clone workflow."