Resource Monitor marked as changed but didn't change, and fails when re-apply it second time

[toni-moreno]

Provider & terraform Version

Terraform v1.4.4
on linux_amd64
+ provider registry.terraform.io/aidanmelen/snowsql v1.3.3
+ provider registry.terraform.io/snowflake-labs/snowflake v0.61.0

Describe the bug

Terraform is marking as changed a snowflake_resource_monitor resource but it didn't change.

Expected behavior

When re-plan without changes it marks resource as changed , when re-apply it fails

Code samples and commands

resource_monitors = {
  "RM_ETL" = {
    quota           = 554
    frequency       = "NEVER"
    start_timestamp = "2023-04-14 12:00"
    end_timestamp   = "2023-11-01 12:00"
    notify_triggers = [75, 100]
    warehouses      = ["ETL"]
  }
}

variable "resource_monitors" {
  description = "lista de resource monitors"
  type = map(object({
    for_account               = optional(bool, false)
    quota                     = number 
    frequency                 = optional(string)
    start_timestamp           = string
    end_timestamp             = string
    notify_triggers           = list(number)
    suspend_trigger           = optional(number) 
    suspend_immediate_trigger = optional(number) 
    notify_users              = optional(list(string))
    warehouses                = optional(set(string))
  }))
 #....some validations here
}


resource "snowflake_resource_monitor" "RM" {

  for_each = var.resource_monitors

  set_for_account = each.value.for_account

  name         = each.key
  credit_quota = each.value.quota

  frequency       = each.value.frequency
  start_timestamp = each.value.start_timestamp
  end_timestamp   = each.value.end_timestamp

  notify_triggers           = each.value.notify_triggers
  suspend_trigger           = each.value.suspend_trigger
  suspend_immediate_trigger = each.value.suspend_immediate_trigger

  notify_users = each.value.notify_users
  warehouses   = each.value.warehouses

  depends_on = [snowflake_warehouse.warehouse]

}

When executing first terraform apply everything seems ok but in snowflake ui the time is not the same than expected form my terraform code

My Snowflake account is in switzerland-north.azure and my timezone is Europe/Madrid.

When re-plan without changes the provider is showing me the resource as changed.

  # snowflake_resource_monitor.RM["RM_ETL"] will be updated in-place
  ~ resource "snowflake_resource_monitor" "RM" {
      ~ end_timestamp             = "2023-11-01T05:00:00-07:00" -> "2023-11-01 12:00"
        id                        = "RM_ETL"
        name                      = "RM_ETL"
      ~ start_timestamp           = "2023-04-14T05:00:00-07:00" -> "2023-04-14 12:00"
        # (7 unchanged attributes hidden)
    }

When re-apply ( force it ) it fails

╷
│ Error: error updating resource monitor RM_ETL
│ 090259 (42601): Must specify frequency and start time together.
│ 
│   with snowflake_resource_monitor.RM["RM_ETL"],
│   on 001_resource_monitors.tf line 5, in resource "snowflake_resource_monitor" "RM":
│    5: resource "snowflake_resource_monitor" "RM" {
│ 
╵

Here there is 3 things to review .

1. what timezone is taking the provider by default
2. how can I change in the end_timestamp and start_timestamp parameters ?
3. ( the bug) why is talking about changed parameters that indeed didn't change
4. (the bug) why is failing when force re-apply.

[toni-moreno]

reviewing the error, seems like TIMESTAMP_OUTPUT_FORMAT is the default YYYY-MM-DD HH24:MI:SS.FF3 TZHTZM and when the provider runs a SHOW RESOURCE MONITORS here:

terraform-provider-snowflake/pkg/snowflake/resource_monitor.go

Lines 261 to 278 in d9f634a

	func ListResourceMonitors(db *sql.DB) ([]ResourceMonitor, error) {
	stmt := "SHOW RESOURCE MONITORS"
	rows, err := Query(db, stmt)
	if err != nil {
	return nil, err
	}
	defer rows.Close()
	dbs := []ResourceMonitor{}
	if err := sqlx.StructScan(rows, &dbs); err != nil {
	if errors.Is(err, sql.ErrNoRows) {
	log.Println("[DEBUG] no resource monitors found")
	return nil, nil
	}
	return nil, fmt.Errorf("unable to scan row for %s err = %w", stmt, err)
	}
	return dbs, nil
	}

It gets the following timestamp

that is different than used in the resource
this is becouse terraform thinks it has changed

# snowflake_resource_monitor.RM["RM_ETL"] will be updated in-place
 ~ resource "snowflake_resource_monitor" "RM" {
     ~ end_timestamp             = "2023-11-01T05:00:00-07:00" -> "2023-11-01 12:00"
       id                        = "RM_ETL"
       name                      = "RM_ETL"
     ~ start_timestamp           = "2023-04-14T05:00:00-07:00" -> "2023-04-14 12:00"
       # (7 unchanged attributes hidden)
   }

But the resource doesn't accept other format than "YYYY-MM-DD HH:MM" as input. when trying to change timestamp to the default output format

"RM_ETL" = {
    quota     = 100
    frequency = "NEVER"
    start_timestamp = "2023-29-03T05:00:00-07:00"
    end_timestamp   = "2023-11-01T05:00:00-07:00"
    notify_triggers = [75, 100]
    warehouses      = ["ETL"]
  }

this error happens.

Invalid date/time format string '2023-29-03T05:00:00-07:00': Invalid format: "2023-29-03T05:00:00-07:00" is malformed at "T05:00:00-07:00"

the solution seems to change TIMESTAMP_OUTPUT_FORMAT to the same format than supported in input before the SHOW RESOURCES MONITOR in the current terraform opened session.

alter session set TIMESTAMP_OUTPUT_FORMAT = 'YYYY-MM-DD HH24:MI'

• There is any way to change the terraform session behavior to handle resource monitor ?

[sfc-gh-jcieslak]

Hello 👋
Please refer to this response - #1175 (comment). I'll let you know whenever the updated resource is available. I think altering the session in the provider currently wouldn't make sense because we are recreating the Snowflake client a few times during the terraform apply, which could make altering session parameters useless. I guess the only way (at least in the current version of the provider) is to use the values proposed by the Snowflake, so they will match during the Read. In the later versions any timestamp will be possible to define.

[sfc-gh-jcieslak]

Hey 👋
The new and refactored resource monitor was released yesterday in version 0.96.0 of the provider. Please migrate with migration guide and let me know if the issue has been resolved so we could close the ticket. Thanks in advance 🙏.

[sfc-gh-jcieslak]

Closing due to long inactivity. Please, create another issue if you think the problem is still not resolved.