Cosmos DB settings
Based on performance testing that was done, the following partition keys and indexes should be defined on the CSM Reachability database.
Recommended Throughput for Bank Directory Plus on Cosmos DB
Recommended settings for Party Entity ingestion for BANKDIRECTORYPLUS full file:
Size of the file: ~1.5 million records
| Config Item | Value | ConfigMap Property |
|---|---|---|
Throughput |
|
Cosmos DB schema |
Direct parallelism |
|
|
Party entity identifiers parallelism |
|
|
Party entity parallelism |
|
|
These values were determined through performance testing, reducing ingestion time from about 175 minutes to about 45 minutes.
Recommended Throughput for Party Entity IDENTIFIERS-ALL file on Cosmos DB
Recommended Throughput for Party Entity ingestion for IDENTIFIERS-ALL full file:
Size of the file: ~4.9 million records, but after some filtering operations, 3 115 384 records are ingested.
| Config Item | Value | ConfigMap Property |
|---|---|---|
Throughput |
|
Cosmos DB schema |
Direct parallelism |
|
|
Party entity identifiers parallelism |
|
|
Party entity parallelism |
|
|
These values were determined through performance testing, with ingestion time ~160 minutes .
For both Bank Directory Plus and Identifiers-ALL, adjust as needed, according to future load testing and business requirements.
Partition Keys
In order to have normal time for processing large files on CosmosDB, we need to add partition keys for every collection. Partition key needs to be added directly on Azure portal. The way to add partition keys:
First we need to open Mongo shell and type use {database_name} e.g.
use ipf-csm-testAfter you have switched to the required database, type commands below:
db.runCommand({customAction:"CreateCollection", collection: "settings-approvals", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-history", shardKey:"_entityId"})
db.runCommand({customAction:"CreateCollection", collection: "settings-agent-settings", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-agent-clearing-settings", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-agent-settlement-settings", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-cs-agent-selection-settings", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-iban-plus", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-iban-structure", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-participant", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-processing-entity", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-party-entity", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-generic-processing-settings", shardKey:"_id"})
db.runCommand({customAction:"CreateCollection", collection: "settings-bicdir2018", shardKey:"_id"})The below indexes are required for the correct operation and optimal performance of the application. They should be created automatically at application startup. If they have not been created automatically (usually an issue that occurs with Cosmos deployments) please ensure the following overrides are in place
ipf.mongodb.max-retry = 6
ipf.mongodb.retryable-error-codes = [16500, 48]
ipf.mongodb.delay-duration = 1000ms
During testing, it was discovered that programmatic index creation on Cosmos would occasionally fail and need an increased number of retries to guarantee success, the above overrides have been tested and confirmed to ensure indexes get created properly.
Typically core indexes are affected since the retry settings are defined in an external module, whereas the application indexes already have the right settings.
Enabling the server-side retry feature on CosmosDB is highly recommended. If for some reason you need to keep this feature disabled, you may need to increase the value of ipf.mongodb.max-retry setting to avoid rate limit errors failing your DB operations.
|