correct connection string

Question

We have two Centera configured as cluster with 4 nodes The nodes have access,replication,management,storage  as NodeRoles configured Each Centera have 4 IP Adresses configured and we have 5 pools Pool 1 to Pool 5 The replication is bi-directional We use Easy archive and the description of the  connection string in the Admin Guide of Easy Archive is not really clear for me. Centera A have the IPs 10.10.10.10, 10.10.10.11,10.10.10.12 , 10.10.10.13 Centera B use 10.10.10.14, 10.10.10.15, 10.10.10.16, 10.10.10.17 I want configure all IP Adresses  from both Centera Systems  in my Storage Provider for best Performance and for highest availability Can I configure the follow connection strings in the Storage Provider ? Is this conform with EMC Centera ? Connection String for Pool1   '10.10.10.10, primary=10.10.10.11, primary=10.10.10.12, primary=10.10.10.13, secondary=10.10.10.14, secondary=10.10.10.15, secondary=10.10.10.16,secondary=10.10.10.17, ?D:\EASY\centera-pea-file\Pool1.pea' Connection String for Pool2 '10.10.10.11, primary=10.10.10.12, primary=10.10.10.13, primary=10.10.10.10, secondary=10.10.10.15, secondary=10.10.10.16, secondary=10.10.10.17,secondary=10.10.10.14, ?D:\EASY\centera-pea-file\Pool2.pea' Connection String for Pool3   '10.10.10.12, primary=10.10.10.13, primary=10.10.10.10, primary=10.10.10.11, secondary=10.10.10.16, secondary=10.10.10.17, secondary=10.10.10.15,secondary=10.10.10.14, ?D:\EASY\centera-pea-file\Pool3.pea'   Connection String for Pool4 '10.10.10.13, primary=10.10.10.10, primary=10.10.10.11, primary=10.10.10.12, secondary=10.10.10.17, secondary=10.10.10.14, secondary=10.10.10.15,secondary=10.10.10.16, ?D:\EASY\centera-pea-file\Pool4.pea' Connection String for Pool5 '10.10.10.10, primary=10.10.10.11, primary=10.10.10.12, primary=10.10.10.13, secondary=10.10.10.14, secondary=10.10.10.15, secondary=10.10.10.16,secondary=10.10.10.17, ?D:\EASY\centera-pea-file\Pool5.pea' what I want to achieve with it is a static round robin I want spead the pools over the four IP's on Centera A, when one IP failed must the Application use the next primary IP Address. When the Centera A is unreacheable must switch the Apllication to the first secondary IP Address (Centera B) Is that possible or what is the best way Thanks alot Andreas

mckeown_paul · Accepted Answer

They are both correct but are not doing what you expect.  Unless you have set up seperate access node networks then both connection strings will effectively use all 4 access nodes on the primary regardless of the IPs you specify in the connection string

mckeown_paul · Answer

Hi there

The Centera SDK will decide on how best to distribute IO across the Access Nodes of the primary Centera automatically.

Note that a secondary centera is a read only centera and not thr replica. The secondary is mean to be say an old centera that has filled up and for some reason the customer doesnt want to expand it and rolls out a new primary but still wants the old now read only centera to be included in reads. New ingest would go to the primary but reads of old content would first be looked for on the primary and then the secondary.

Best practise is to put at least two IP addresses of the primary access nodes (plus any PEA file). The SDk will attempt to connect to the IP on the left and if successfull it will then probe that node for the IP addresses of all available access nodes in the primary and make connections to them. It will then load balance IO across all available primary Access Nodes.

The SDK will also discover the IP addresses of all the replica access nodes and may make connections to them (default is yes but you can disable by setting lazy pool open). If a read fails on the primary by default the sdk will ask the replica if it has the clip and read it from there.

You used to not be able to segregate pool traffic to particular Access Nodes however you can from rev 4.2 of Centrastar create and manage groups of access nodes (min two max half of available) and have them used by seperate applications (or instances of an application). However I wouldn't recommend using it unless you are rolling out classes of service to internal or external users. Just let the SDK load balance across all available access nodes.

mckeown_paul · Answer

I should add that the use of the keyword primary is uneccessary.

andreasschulz1 · Answer

Thanks for fast reply,

in my case is this correct ?

Connection String for Pool1

“10.10.10.10, primary=10.10.10.11 ?D:\EASY\centera-pea-file\Pool1.pea”

Connection String for Pool2

“10.10.10.12, primary=10.10.10.13 ?D:\EASY\centera-pea-file\Pool2.pea”

Thanks

Andreas

andreasschulz1 · Answer

Thanks alot, I will set '10.10.10.10,10.10.10.11 ?D:\EASY\centera-pea-file\Pool1.pea' for all pools Best Regards Andreas

Centera

Was this post helpful?