RDB-Loader Redshift - Invalid operation: column "filters" duplicated

I setup the Snowplow for Opensource pipeline in AWS using the quick examples, and added a “transformer-kinesis-ec2/aws” instance to shred the enriched messages, plus an instance running the snowplow/rdb-loader-redshift:5.1.1-rc2 to send the events to my Redshift cluster.
Now the RDB loader for Redshift is able to copy to the database all the events I tried so far except the one generated by the trackSiteSearch method. It fails when creating the “atomic.com_snowplowanalytics_snowplow_site_search_1” table with the following error:

rdb-loader-redshift_1  | INFO com.snowplowanalytics.snowplow.rdbloader: RDB Loader 5.1.1-rc2 has started.
rdb-loader-redshift_1  | INFO HikariPool-1 - Starting...
rdb-loader-redshift_1  | INFO HikariPool-1 - Driver does not support get/set network timeout for connections. ([Amazon][JDBC](10220) Driver does not support this optional feature.)
rdb-loader-redshift_1  | INFO HikariPool-1 - Start completed.
rdb-loader-redshift_1  | INFO Loader: Target check is completed
rdb-loader-redshift_1  | INFO Loader: No operation prepare step is completed
rdb-loader-redshift_1  | INFO Manifest: No changes needed on the manifest table
rdb-loader-redshift_1  | INFO Loader: Manifest initialization is completed
rdb-loader-redshift_1  | INFO Loader: load_tstamp column already exists
rdb-loader-redshift_1  | INFO Loader: Adding load_tstamp column is completed
rdb-loader-redshift_1  | INFO FolderMonitoring: Configuration for monitoring.folders hasn't been provided - monitoring is disabled
rdb-loader-redshift_1  | INFO DataDiscovery: Received a new message
rdb-loader-redshift_1  | INFO DataDiscovery: Total 1 messages received, 0 loaded
rdb-loader-redshift_1  | INFO DataDiscovery: New data discovery at run=2022-11-08-22-25-00 with following shredded types:
rdb-loader-redshift_1  |   * iglu:com.snowplowanalytics.snowplow/web_page/jsonschema/1-*-* TSV
rdb-loader-redshift_1  |   * iglu:nl.basjes/yauaa_context/jsonschema/1-*-* TSV
rdb-loader-redshift_1  |   * iglu:com.google.analytics.enhanced-ecommerce/impressionFieldObject/jsonschema/1-*-* TSV
rdb-loader-redshift_1  |   * iglu:com.snowplowanalytics.snowplow/site_search/jsonschema/1-*-* TSV
rdb-loader-redshift_1  |   * iglu:com.snowplowanalytics.snowplow/ua_parser_context/jsonschema/1-*-* TSV
rdb-loader-redshift_1  |   * iglu:com.google.analytics.enhanced-ecommerce/action/jsonschema/1-*-* TSV
rdb-loader-redshift_1  | INFO Load: Loading transaction for s3://snowplow-vui-poc/transformed/good/run=2022-11-08-22-25-00/ has started
rdb-loader-redshift_1  | INFO Migration: Creating atomic.com_snowplowanalytics_snowplow_site_search_1 table for iglu:com.snowplowanalytics.snowplow/site_search/jsonschema/1-0-0
rdb-loader-redshift_1  | ERROR Loader: Loading of s3://snowplow-vui-poc/transformed/good/run=2022-11-08-22-25-00/ has failed. Not adding into retry queue. [Amazon](500310) Invalid operation: column "filters" duplicated; - SqlState: 42701
rdb-loader-redshift_1  | INFO Acking SQS message because processing is complete.

The problem is I believe caused by the transformer which adds a duplicated “filters” column in the shredded TSV file:

com.snowplowanalytics.snowplow	site_search	jsonschema	1-0-0	e7e52081-1d66-4dde-9351-7f03e655cd87	2022-11-08 22:14:48.189	events	["events","site_search"]	events	["bananas"]	{"category":"books","sub-category":"non-fiction"}	{"category":"books","sub-category":"non-fiction"}	48	48

I don’t understand how the transformer is adding the “filters” column twice. Can anybody shred some light on this problem?
My transformer terraform configuration is the following:

module "transformer_enriched_redshift" {
  source = "snowplow-devops/transformer-kinesis-ec2/aws"
  version = "0.1.0"

  count = 1

  name                           = "${var.prefix}-transformer-kinesis-enriched-server"
  vpc_id                         = var.vpc_id
  subnet_ids                     = var.public_subnet_ids
  ssh_key_name                   = aws_key_pair.pipeline.key_name
  ssh_ip_allowlist               = var.ssh_ip_allowlist
  stream_name                    = module.enriched_stream.name
  s3_bucket_name                 = var.s3_bucket_name
  s3_bucket_object_prefix        = "${var.s3_bucket_object_prefix}transformed/good"
  window_period_min              = var.transformer_window_period_min
  sqs_queue_name                 = aws_sqs_queue.message_queue_rs[0].name
  transformation_type            = "shred"
  custom_iglu_resolvers          = local.custom_iglu_resolvers
  kcl_write_max_capacity         = var.pipeline_kcl_write_max_capacity
  iam_permissions_boundary       = var.iam_permissions_boundary
  telemetry_enabled              = var.telemetry_enabled
  user_provided_id               = var.user_provided_id
  tags                           = var.tags
  cloudwatch_logs_enabled        = var.cloudwatch_logs_enabled
  cloudwatch_logs_retention_days = var.cloudwatch_logs_retention_days

resource "aws_sqs_queue" "message_queue_rs" {
  count                       =  1
  content_based_deduplication = true
  name                        = "${var.prefix}-sf-loader.fifo"
  fifo_queue                  = true
  kms_master_key_id           = "alias/aws/sqs"

Thanks in advance.

Hi @dgiulianelli this looks like it’s a bug in the transformer/loader. I have opened an issue in Github and I’ve got a preliminary fix for it already. We are planning a new release of RDB Loader fairly soon, so I reckon we can get this one fixed for you pretty quickly.

Sorry for this inconvenience. I will update this thread once the fix has been released.

1 Like

Great! Thank you @istreeter for the prompt response and for quickly identifying the issue. I really appreciate it. Looking forward to hear from you with the official fix.

@istreeter , any updates on this issue? Do you have an official release of the RDB Loader with this fix?
Thanks in advance.

Hi @dgiulianelli, RDB Loader 5.2.1 should fix this issue. Let us know if you encounter with any other problem.

1 Like

Thank you very much @enes_aldemir. I just enabled version 5.2.1 for both the transformer and the RDB loader and now I can finally see the com_snowplowanalytics_snowplow_site_search_1 table populated in the atomic schema of my Redshift cluster.
I really appreciate you guys fixing this issue in a short amount of time.

1 Like