No public description

PiperOrigin-RevId: 659608688

Author: sjswerdlow

modules/sap_hana/README.md

@@ -49,8 +49,11 @@ https://cloud.google.com/docs/terraform
 | data\_disk\_size\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Overrides the default size for the data disk(s), that is based off of the machine\_type. | `number` | `null` | no |
 | data\_disk\_throughput\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Directly sets the throughput in MB/s that the data disk(s) will use. Has no effect if not using a disk type that supports it. | `number` | `null` | no |
 | data\_disk\_type\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Override the 'default\_disk\_type' for the data disk. | `string` | `""` | no |
+| data\_stripe\_size | Optional - default is 256k. Stripe size for data volume striping (if enable\_data\_striping = true). | `string` | `"256k"` | no |
 | disk\_type | Optional - The default disk type to use for disk(s) containing log and data volumes. The default is pd-ssd, except for machines that do not support PD, in which case the default is hyperdisk-extreme. Not all disk are supported on all machine types - see https://cloud.google.com/compute/docs/disks/ for details. | `string` | `""` | no |
+| enable\_data\_striping | Optional - default is false. Enable LVM striping of data volume across multiple disks. | `bool` | `false` | no |
 | enable\_fast\_restart | Optional - The default is true. If set enables HANA Fast Restart. | `bool` | `true` | no |
+| enable\_log\_striping | Optional - default is false. Enable LVM striping of log volume across multiple disks. | `bool` | `false` | no |
 | hyperdisk\_balanced\_iops\_default | Optional - default is 3000. Number of IOPS that is set for each disk of type Hyperdisk-balanced (except for boot/usrsap/shared disks). | `number` | `3000` | no |
 | hyperdisk\_balanced\_throughput\_default | Optional - default is 750. Throughput in MB/s that is set for each disk of type Hyperdisk-balanced (except for boot/usrsap/shared disks). | `number` | `750` | no |
 | include\_backup\_disk | Optional - The default is true. If set creates a disk for backups. | `bool` | `true` | no |
@@ -61,9 +64,12 @@ https://cloud.google.com/docs/terraform
 | log\_disk\_size\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Overrides the default size for the log disk(s), that is based off of the machine\_type. | `number` | `null` | no |
 | log\_disk\_throughput\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Directly sets the throughput in MB/s that the log disk(s) will use. Has no effect if not using a disk type that supports it. | `number` | `null` | no |
 | log\_disk\_type\_override | Warning, do not use unless instructed or aware of the implications of using this setting. Override the 'default\_disk\_type' for the log disk. | `string` | `""` | no |
+| log\_stripe\_size | Optional - default is 64k. Stripe size for log volume striping (if enable\_log\_striping = true). | `string` | `"64k"` | no |
 | machine\_type | Machine type for the instances. | `string` | n/a | yes |
 | network\_tags | OPTIONAL - Network tags can be associated to your instance on deployment. This can be used for firewalling or routing purposes. | `list(string)` | `[]` | no |
 | nic\_type | Optional - This value determines the type of NIC to use, valid options are GVNIC and VIRTIO\_NET. If choosing GVNIC make sure that it is supported by your OS choice here https://cloud.google.com/compute/docs/images/os-details#networking. | `string` | `""` | no |
+| number\_data\_disks | Optional - default is 2. Number of disks to use for data volume striping (if enable\_data\_striping = true). | `number` | `2` | no |
+| number\_log\_disks | Optional - default is 2. Number of disks to use for log volume striping (if enable\_log\_striping = true). | `number` | `2` | no |
 | post\_deployment\_script | OPTIONAL - gs:// or https:// location of a script to execute on the created VM's post deployment. | `string` | `""` | no |
 | primary\_startup\_url | Startup script to be executed when the VM boots, should not be overridden. | `string` | `"curl -s https://storage.googleapis.com/cloudsapdeploy/terraform/latest/terraform/sap_hana/hana_startup.sh | bash -s https://storage.googleapis.com/cloudsapdeploy/terraform/latest/terraform"` | no |
 | project\_id | Project id where the instances will be created. | `string` | n/a | yes |

modules/sap_hana/main.tf

@@ -80,6 +80,8 @@ locals {
   default_hyperdisk_extreme  = (length(regexall("^x4-", var.machine_type)) > 0)
   default_hyperdisk_balanced = (length(regexall("^c4-|^c3-.*-metal", var.machine_type)) > 0)
   only_hyperdisks_supported  = local.default_hyperdisk_extreme || local.default_hyperdisk_balanced
+  num_data_disks             = var.enable_data_striping ? var.number_data_disks : 1
+  num_log_disks              = var.enable_log_striping ? var.number_log_disks : 1
 
   # Minimum disk sizes are used to ensure throughput. Extreme disks don't need this.
   # All 'over provisioned' capacity is to go onto the data disk.
@@ -113,10 +115,11 @@ locals {
 
   unified_pd_size        = var.unified_disk_size_override == null ? local.pd_size : var.unified_disk_size_override
   unified_worker_pd_size = var.unified_worker_disk_size_override == null ? local.pd_size_worker : var.unified_worker_disk_size_override
-  data_pd_size           = var.data_disk_size_override == null ? local.hana_data_size : var.data_disk_size_override
-  log_pd_size            = var.log_disk_size_override == null ? local.hana_log_size : var.log_disk_size_override
-  shared_pd_size         = var.shared_disk_size_override == null ? local.hana_shared_size : var.shared_disk_size_override
-  usrsap_pd_size         = var.usrsap_disk_size_override == null ? local.hana_usrsap_size : var.usrsap_disk_size_override
+  # for striping: divide data/log size by number of disks
+  data_pd_size   = var.data_disk_size_override == null ? ceil(local.hana_data_size / local.num_data_disks) : ceil(var.data_disk_size_override / local.num_data_disks)
+  log_pd_size    = var.log_disk_size_override == null ? ceil(local.hana_log_size / local.num_log_disks) : ceil(var.log_disk_size_override / local.num_log_disks)
+  shared_pd_size = var.shared_disk_size_override == null ? local.hana_shared_size : var.shared_disk_size_override
+  usrsap_pd_size = var.usrsap_disk_size_override == null ? local.hana_usrsap_size : var.usrsap_disk_size_override
 
   # Disk types
   final_data_disk_type   = var.data_disk_type_override == "" ? local.final_disk_type : var.data_disk_type_override
@@ -129,12 +132,12 @@ locals {
 
   # Disk IOPS
   hdx_iops_map = {
-    "data"    = max(10000, local.data_pd_size * 2)
-    "log"     = max(10000, local.log_pd_size * 2)
+    "data"    = max(10000, local.data_pd_size * 2 * local.num_data_disks)
+    "log"     = max(10000, local.log_pd_size * 2 * local.num_log_disks)
     "shared"  = null
     "usrsap"  = null
-    "unified" = max(10000, local.data_pd_size * 2) + max(10000, local.log_pd_size * 2)
-    "worker"  = max(10000, local.data_pd_size * 2) + max(10000, local.log_pd_size * 2)
+    "unified" = max(10000, local.data_pd_size * 2 * local.num_data_disks) + max(10000, local.log_pd_size * 2 * local.num_log_disks)
+    "worker"  = max(10000, local.data_pd_size * 2 * local.num_data_disks) + max(10000, local.log_pd_size * 2 * local.num_log_disks)
     "backup"  = max(10000, 2 * local.backup_size)
   }
   hdb_iops_map = {
@@ -163,8 +166,13 @@ locals {
     "hyperdisk-extreme"  = local.hdx_iops_map
   }
 
-  final_data_iops           = var.data_disk_iops_override == null ? local.iops_map[local.final_data_disk_type]["data"] : var.data_disk_iops_override
-  final_log_iops            = var.log_disk_iops_override == null ? local.iops_map[local.final_log_disk_type]["log"] : var.log_disk_iops_override
+  # for striping: divide data/log IOPS by number of disks
+  final_data_iops = (var.data_disk_iops_override == null ?
+    (local.iops_map[local.final_data_disk_type]["data"] == null ? null : ceil(local.iops_map[local.final_data_disk_type]["data"] / local.num_data_disks)
+  ) : ceil(var.data_disk_iops_override / local.num_data_disks))
+  final_log_iops = (var.log_disk_iops_override == null ?
+    (local.iops_map[local.final_log_disk_type]["log"] == null ? null : ceil(local.iops_map[local.final_log_disk_type]["log"] / local.num_log_disks)
+  ) : ceil(var.log_disk_iops_override / local.num_log_disks))
   final_shared_iops         = var.shared_disk_iops_override == null ? local.iops_map[local.final_shared_disk_type]["shared"] : var.shared_disk_iops_override
   final_usrsap_iops         = var.usrsap_disk_iops_override == null ? local.iops_map[local.final_usrsap_disk_type]["usrsap"] : var.usrsap_disk_iops_override
   final_unified_iops        = var.unified_disk_iops_override == null ? local.iops_map[local.final_disk_type]["unified"] : var.unified_disk_iops_override
@@ -198,8 +206,13 @@ locals {
     "hyperdisk-extreme"  = local.null_throughput_map
   }
 
-  final_data_throughput           = var.data_disk_throughput_override == null ? local.throughput_map[local.final_data_disk_type]["data"] : var.data_disk_throughput_override
-  final_log_throughput            = var.log_disk_throughput_override == null ? local.throughput_map[local.final_log_disk_type]["log"] : var.log_disk_throughput_override
+  # for striping: divide throughput by number of disks
+  final_data_throughput = (var.data_disk_throughput_override == null ?
+    (local.throughput_map[local.final_data_disk_type]["data"] == null ? null : ceil(local.throughput_map[local.final_data_disk_type]["data"] / local.num_data_disks)
+  ) : ceil(var.data_disk_throughput_override / local.num_data_disks))
+  final_log_throughput = (var.log_disk_throughput_override == null ?
+    (local.throughput_map[local.final_log_disk_type]["log"] == null ? null : ceil(local.throughput_map[local.final_log_disk_type]["log"] / local.num_log_disks)
+  ) : ceil(var.log_disk_throughput_override / local.num_log_disks))
   final_shared_throughput         = var.shared_disk_throughput_override == null ? local.throughput_map[local.final_shared_disk_type]["shared"] : var.shared_disk_throughput_override
   final_usrsap_throughput         = var.usrsap_disk_throughput_override == null ? local.throughput_map[local.final_usrsap_disk_type]["usrsap"] : var.usrsap_disk_throughput_override
   final_unified_throughput        = var.unified_disk_throughput_override == null ? local.throughput_map[local.final_disk_type]["unified"] : var.unified_disk_throughput_override
@@ -254,6 +267,21 @@ data "assert_test" "verify_hyperdisk_usage_for_backup_disk" {
   test  = local.only_hyperdisks_supported && local.use_backup_disk ? (length(regexall("hyperdisk", local.final_backup_disk_type)) > 0) : true
   throw = "The selected a machine type only works with hyperdisks. Set 'backup_disk_type' accordingly, e.g. 'backup_disk_type = hyperdisk-balanced'"
 }
+# tflint-ignore: terraform_unused_declarations
+data "assert_test" "striping_with_split_disk" {
+  test  = ((var.enable_data_striping || var.enable_log_striping) && !var.use_single_shared_data_log_disk) || !(var.enable_data_striping || var.enable_log_striping)
+  throw = "Striping not supported if log and data are on the same disk(s). To use striping set 'use_single_shared_data_log_disk=false'"
+}
+# tflint-ignore: terraform_unused_declarations
+data "validation_warning" "warn_data_striping" {
+  condition = var.enable_data_striping
+  summary   = "Data striping is only intended for cases where the machine level limits are higher than the hyperdisk disk level limits. Refer to https://cloud.google.com/compute/docs/disks/hyperdisks#hd-performance-limits"
+}
+# tflint-ignore: terraform_unused_declarations
+data "validation_warning" "warn_log_striping" {
+  condition = var.enable_log_striping
+  summary   = "Log striping is not a recommended deployment option."
+}
 
 ################################################################################
 # disks
@@ -298,20 +326,25 @@ resource "google_compute_disk" "sap_hana_unified_worker_disks" {
 }
 
 # Split data/log/sap disks
+#  name without striping:  00001, 00002, ...
+#  name with striping: 00001-01, 00001-02, 00001-03, 00002-01, 00002-02, 00002-03, ...
 resource "google_compute_disk" "sap_hana_data_disks" {
-  count                  = var.use_single_shared_data_log_disk ? 0 : var.sap_hana_scaleout_nodes + 1
-  name                   = format("${var.instance_name}-data%05d", count.index + 1)
+  count = var.use_single_shared_data_log_disk ? 0 : (var.sap_hana_scaleout_nodes + 1) * local.num_data_disks
+  name = (var.enable_data_striping ?
+    format("${var.instance_name}-data%05d-%02d", floor(count.index / local.num_data_disks) + 1, (count.index % local.num_data_disks) + 1) :
+  format("${var.instance_name}-data%05d", count.index + 1))
   type                   = local.final_data_disk_type
   zone                   = var.zone
   size                   = local.data_pd_size
   project                = var.project_id
   provisioned_iops       = local.final_data_iops
   provisioned_throughput = local.final_data_throughput
 }
-
 resource "google_compute_disk" "sap_hana_log_disks" {
-  count                  = var.use_single_shared_data_log_disk ? 0 : var.sap_hana_scaleout_nodes + 1
-  name                   = format("${var.instance_name}-log%05d", count.index + 1)
+  count = var.use_single_shared_data_log_disk ? 0 : (var.sap_hana_scaleout_nodes + 1) * local.num_log_disks
+  name = (var.enable_log_striping ?
+    format("${var.instance_name}-log%05d-%02d", floor(count.index / local.num_log_disks) + 1, (count.index % local.num_log_disks) + 1) :
+  format("${var.instance_name}-log%05d", count.index + 1))
   type                   = local.final_log_disk_type
   zone                   = var.zone
   size                   = local.log_pd_size
@@ -405,19 +438,22 @@ resource "google_compute_instance" "sap_hana_primary_instance" {
       source      = google_compute_disk.sap_hana_unified_disks[0].self_link
     }
   }
-
   dynamic "attached_disk" {
-    for_each = var.use_single_shared_data_log_disk ? [] : [1]
+    for_each = var.use_single_shared_data_log_disk ? [] : [for i in range(local.num_data_disks) : {
+      final_disk_index = i
+    }]
     content {
-      device_name = google_compute_disk.sap_hana_data_disks[0].name
-      source      = google_compute_disk.sap_hana_data_disks[0].self_link
+      device_name = google_compute_disk.sap_hana_data_disks[attached_disk.value.final_disk_index].name
+      source      = google_compute_disk.sap_hana_data_disks[attached_disk.value.final_disk_index].self_link
     }
   }
   dynamic "attached_disk" {
-    for_each = var.use_single_shared_data_log_disk ? [] : [1]
+    for_each = var.use_single_shared_data_log_disk ? [] : [for i in range(local.num_log_disks) : {
+      final_disk_index = i
+    }]
     content {
-      device_name = google_compute_disk.sap_hana_log_disks[0].name
-      source      = google_compute_disk.sap_hana_log_disks[0].self_link
+      device_name = google_compute_disk.sap_hana_log_disks[attached_disk.value.final_disk_index].name
+      source      = google_compute_disk.sap_hana_log_disks[attached_disk.value.final_disk_index].self_link
     }
   }
   dynamic "attached_disk" {
@@ -501,6 +537,8 @@ resource "google_compute_instance" "sap_hana_primary_instance" {
     sap_hana_data_disk_type         = local.final_data_disk_type
     enable_fast_restart             = var.enable_fast_restart
     native_bm                       = local.native_bm
+    data_stripe_size                = var.data_stripe_size
+    log_stripe_size                 = var.log_stripe_size
     template-type                   = "TERRAFORM"
   }
 
@@ -539,19 +577,22 @@ resource "google_compute_instance" "sap_hana_worker_instances" {
       source      = google_compute_disk.sap_hana_unified_worker_disks[count.index].self_link
     }
   }
-
   dynamic "attached_disk" {
-    for_each = var.use_single_shared_data_log_disk ? [] : [1]
+    for_each = var.use_single_shared_data_log_disk ? [] : [for i in range(local.num_data_disks) : {
+      final_disk_index = i + (count.index + 1) * local.num_data_disks
+    }]
     content {
-      device_name = google_compute_disk.sap_hana_data_disks[count.index + 1].name
-      source      = google_compute_disk.sap_hana_data_disks[count.index + 1].self_link
+      device_name = google_compute_disk.sap_hana_data_disks[attached_disk.value.final_disk_index].name
+      source      = google_compute_disk.sap_hana_data_disks[attached_disk.value.final_disk_index].self_link
     }
   }
   dynamic "attached_disk" {
-    for_each = var.use_single_shared_data_log_disk ? [] : [1]
+    for_each = var.use_single_shared_data_log_disk ? [] : [for i in range(local.num_log_disks) : {
+      final_disk_index = i + (count.index + 1) * local.num_log_disks
+    }]
     content {
-      device_name = google_compute_disk.sap_hana_log_disks[count.index + 1].name
-      source      = google_compute_disk.sap_hana_log_disks[count.index + 1].self_link
+      device_name = google_compute_disk.sap_hana_log_disks[attached_disk.value.final_disk_index].name
+      source      = google_compute_disk.sap_hana_log_disks[attached_disk.value.final_disk_index].self_link
     }
   }
   dynamic "attached_disk" {
@@ -619,6 +660,8 @@ resource "google_compute_instance" "sap_hana_worker_instances" {
     sap_hana_shared_disk            = false
     enable_fast_restart             = var.enable_fast_restart
     native_bm                       = local.native_bm
+    data_stripe_size                = var.data_stripe_size
+    log_stripe_size                 = var.log_stripe_size
     template-type                   = "TERRAFORM"
   }
 

modules/sap_hana/sap_hana.tf

@@ -76,4 +76,5 @@ module "sap_hana" {
   # include_backup_disk             = true_or_false         # default is true
   # backup_disk_type                = "DISK_TYPE"           # default is pd-ssd, except for machines that do not support PD, in which case the default is hyperdisk-extreme. Valid types are "pd-ssd", "pd-balanced", "pd-standard", "pd-extreme", "hyperdisk-balanced", "hyperdisk-extreme".
   # enable_fast_restart             = true_or_false         # default is true, whether to enable HANA Fast Restart
+  # enable_data_striping            = true_or_false         # default is false. Enable LVM striping of data volume across multiple disks. Data striping is only intended for cases where the machine level limits are higher than the hyperdisk disk level limits. Refer to https://cloud.google.com/compute/docs/disks/hyperdisks#hd-performance-limits
 }

modules/sap_hana/variables.tf

@@ -338,6 +338,12 @@ variable "hyperdisk_balanced_throughput_default" {
   default     = 750
 }
 
+variable "enable_data_striping" {
+  type        = bool
+  description = "Optional - default is false. Enable LVM striping of data volume across multiple disks."
+  default     = false
+}
+
 #
 # DO NOT MODIFY unless instructed or aware of the implications of using those settings
 #
@@ -497,3 +503,33 @@ variable "can_ip_forward" {
   description = "Whether sending and receiving of packets with non-matching source or destination IPs is allowed."
   default     = true
 }
+
+variable "enable_log_striping" {
+  type        = bool
+  description = "Optional - default is false. Enable LVM striping of log volume across multiple disks."
+  default     = false
+}
+
+variable "number_data_disks" {
+  type        = number
+  description = "Optional - default is 2. Number of disks to use for data volume striping (if enable_data_striping = true)."
+  default     = 2
+}
+
+variable "number_log_disks" {
+  type        = number
+  description = "Optional - default is 2. Number of disks to use for log volume striping (if enable_log_striping = true)."
+  default     = 2
+}
+
+variable "data_stripe_size" {
+  type        = string
+  description = "Optional - default is 256k. Stripe size for data volume striping (if enable_data_striping = true)."
+  default     = "256k"
+}
+
+variable "log_stripe_size" {
+  type        = string
+  description = "Optional - default is 64k. Stripe size for log volume striping (if enable_log_striping = true)."
+  default     = "64k"
+}