Added Spanner Migration Tool alerting and monitoring code for sharded migrations #2017
Conversation
There was a problem hiding this comment.
Thanks for the PR. This is great work. Added some comments, please take a look.
Overall -
Please look at the reference of an existing template and follow a similar structure wherever possible. We also have a status check which validates terraform templates and should ideally validate a bunch of stuff as well, please see it's output.
| @@ -0,0 +1,48 @@ | |||
| # Terraform state files | |||
There was a problem hiding this comment.
Do we need this? I think there is a global gitignore that we can use instead?
There was a problem hiding this comment.
Agree, we do not need this. I have removed this file in the new commit.
| @@ -0,0 +1,63 @@ | |||
| # Spanner Migration Tool Monitoring Dashboard - Terraform Module | |||
There was a problem hiding this comment.
Where possible, can you mirror the sample structure we have for other templates?
Example: We include a terraform.tfvars and terraform_simple.tfvars for each sample. Definitions are here - https://github.com/GoogleCloudPlatform/DataflowTemplates/tree/main/v2/datastream-to-spanner/terraform/samples#sample-structure
As much as possible, please try to keep the naming convention consistent.
There was a problem hiding this comment.
Instead of using "Spanner Migration Tool" as a way to refer, could you please replace it with "Live migration template(s)".
For example -
- Live Migration Monitoring Dashboard (Title)
visualize key metrics related to the Live migration template(s)(first line)
There was a problem hiding this comment.
Updated the README.md and the dashboard title to use Live Migration Monitoring Dashboard
| name = "dataflow_conversion_errors_metric" | ||
|
|
||
| # Filter to capture only conversion errors with severity of ERROR or higher for Dataflow jobs | ||
| filter = "resource.type=\"dataflow_job\" AND severity>=ERROR AND textPayload:\"conversion error\"" |
There was a problem hiding this comment.
@shreyakhajanchi Please validate with the recent updates to the metrics if the filter is still accurate.
| resource "google_logging_metric" "dataflow_conversion_errors_metric" { | ||
| name = "dataflow_conversion_errors_metric" | ||
|
|
||
| # Filter to capture only conversion errors with severity of ERROR or higher for Dataflow jobs | ||
| filter = "resource.type=\"dataflow_job\" AND severity>=ERROR AND textPayload:\"conversion error\"" | ||
|
|
||
| # Metric descriptor settings to define the metric type and value type | ||
| metric_descriptor { | ||
| metric_kind = "DELTA" # Tracks change over time | ||
| value_type = "INT64" # Counts integer values | ||
| } | ||
|
|
||
| } | ||
|
|
||
| # Log-based metric for other (non-conversion) Dataflow errors | ||
| resource "google_logging_metric" "dataflow_other_errors_metric" { | ||
| name = "dataflow_other_errors_metric" | ||
| description = "Metric to track other Dataflow errors (excluding conversion errors)" | ||
|
|
||
| # Filter to capture all Dataflow errors except conversion errors | ||
| filter = "resource.type=\"dataflow_job\" AND severity>=ERROR AND NOT textPayload:\"conversion error\"" | ||
|
|
||
| metric_descriptor { | ||
| metric_kind = "DELTA" | ||
| value_type = "INT64" | ||
| } | ||
|
|
||
| } | ||
|
|
||
| # Log-based metric for total Dataflow errors (both conversion and other errors) | ||
| resource "google_logging_metric" "dataflow_total_errors_metric" { | ||
| name = "dataflow_total_errors_metric" | ||
| description = "Metric to track the total number of Dataflow errors" | ||
|
|
||
| # Filter to capture all Dataflow errors with severity of ERROR or higher | ||
| filter = "resource.type=\"dataflow_job\" AND severity>=ERROR" | ||
|
|
||
| metric_descriptor { | ||
| metric_kind = "DELTA" | ||
| value_type = "INT64" | ||
| } | ||
|
|
||
| } |
There was a problem hiding this comment.
Why do we need log-based metrics? Aren't these metrics directly published to GCP in a counter?
An older (but similar) example of querying Retryable Errors from Monitoring -
fetch dataflow_job
| metric 'dataflow.googleapis.com/job/user_counter'
| filter (resource.job_name == 'hb-dataflow-polltest-100tables-ef7e-3589')
| filter (metric.metric_name == 'Retryable errors')
| group_by 1m, [value_user_counter_mean: mean(value.user_counter)]
| every 1m
|
Sample Status check run - This shows that the file is not formatted with |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #2017 +/- ##
============================================
+ Coverage 45.41% 52.93% +7.51%
+ Complexity 3675 1367 -2308
============================================
Files 842 378 -464
Lines 49970 20661 -29309
Branches 5261 2090 -3171
============================================
- Hits 22692 10936 -11756
+ Misses 25608 9045 -16563
+ Partials 1670 680 -990
|
| pubsub_age_of_oldest_message_threshold = 120 # Maximum age of the oldest message in Pub/Sub in seconds. | ||
|
|
||
| # Notification Configuration | ||
| email_address = "nehamodgil@google.com" # Email address to receive alert notifications. |
| @@ -0,0 +1,63 @@ | |||
| # Live Migration Monitoring Dashboard - Terraform Module | |||
There was a problem hiding this comment.
Please add this template to the list of examples here -
https://github.com/GoogleCloudPlatform/DataflowTemplates/tree/main/v2/datastream-to-spanner/terraform/samples#list-of-examples
|
Looks like there are some formatting issues - Please run |
This Terraform module delivers the following features tailored for the Spanner Migration Tool used in sharded migrations: