Merge "Defining fs.aio-max-nr for hosts with libvirt service" into stable/train

2022-03-07 20:15:42 +00:00 · 2022-03-07 20:15:42 +00:00 · 10e6ce2108
parent c734fef1dc 8c2eaf949a
commit 10e6ce2108
25 changed files with 52 additions and 0 deletions
--- a/deployment/kernel/kernel-baremetal-ansible.yaml
+++ b/deployment/kernel/kernel-baremetal-ansible.yaml
@ -102,9 +102,26 @@ parameters:
    type: json
    tags:
      - role_specific
+  FsAioMaxNumber:
+    default: 0
+    description: >
+      The kernel allocates aio memory on demand, and this number limits the
+      number of parallel aio requests; the only drawback of a larger limit is
+      that a malicious guest could issue parallel requests to cause the kernel
+      to set aside memory.  Set this number at least as large as
+        128 * (number of virtual disks on the host)
+      Libvirt uses a default of 1M requests to allow 8k disks, with at most
+      64M of kernel memory if all disks hit an aio request at the same time.
+    type: number
+    tags:
+      - role_specific

 conditions:
  ipv6_disabled: {equals: [{get_param: KernelDisableIPv6}, 1]}
+  fs_aio_max_number_set:
+    and:
+      - not: {equals: [{get_param: [RoleParameters, FsAioMaxNumber]}, 0]}
+      - not: {equals: [{get_param: [RoleParameters, FsAioMaxNumber]}, '']}

 resources:
  # Merging role-specific parameters (RoleParameters) with the default parameters.
@ -173,4 +190,9 @@ outputs:
                value: {get_param: BridgeNfCallIp6Tables}
              fs.inotify.max_user_instances:
                value: {get_param: InotifyIntancesMax}
+            - if:
+                - fs_aio_max_number_set
+                - fs.aio-max-nr:
+                    value: {get_param: [RoleParameters, FsAioMaxNumber]}
+                - {}
            - {get_attr: [RoleParametersValue, value, extra_sysctl_settings]}
--- a/releasenotes/notes/fixing-fs.aio-max.nr-on-compute-06447122b8e12c71.yaml
+++ b/releasenotes/notes/fixing-fs.aio-max.nr-on-compute-06447122b8e12c71.yaml
@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    When we install libvirt on a host, the system parameter ``fs.aio-max-nr`` is
+    to 1048576. Since we containerized libvirtd, we lost this system parameter.
+    We now make sure it's defined by adding it from the nova-libvirt-common
+    template.
--- a/roles/Compute.yaml
+++ b/roles/Compute.yaml
@ -17,6 +17,7 @@
      subnet: storage_subnet
  HostnameFormatDefault: '%stackname%-novacompute-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  # Deprecated & backward-compatible values (FIXME: Make parameters consistent)
  # Set uses_deprecated_params to True if any deprecated params are used.
--- a/roles/ComputeAlt.yaml
+++ b/roles/ComputeAlt.yaml
@ -14,6 +14,7 @@
      subnet: storage_subnet
  HostnameFormatDefault: '%stackname%-novacomputealt-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeDVR.yaml
+++ b/roles/ComputeDVR.yaml
@ -16,6 +16,7 @@
      subnet: storage_subnet
  HostnameFormatDefault: '%stackname%-novacompute-dvr-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeHCI.yaml
+++ b/roles/ComputeHCI.yaml
@ -14,6 +14,7 @@
    StorageMgmt:
      subnet: storage_mgmt_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "throughput-performance"
  # CephOSD present so serial has to be 1
  update_serial: 1
--- a/roles/ComputeHCIOvsDpdk.yaml
+++ b/roles/ComputeHCIOvsDpdk.yaml
@ -16,6 +16,7 @@
  # CephOSD present so serial has to be 1
  update_serial: 1
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "cpu-partitioning"
    VhostuserSocketGroup: "hugetlbfs"
    NovaLibvirtRxQueueSize: 1024
--- a/roles/ComputeInstanceHA.yaml
+++ b/roles/ComputeInstanceHA.yaml
@ -14,6 +14,7 @@
      subnet: storage_subnet
  HostnameFormatDefault: '%stackname%-novacomputeiha-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeLocalEphemeral.yaml
+++ b/roles/ComputeLocalEphemeral.yaml
@ -18,6 +18,7 @@
  update_serial: 25
  HostnameFormatDefault: '%stackname%-novacompute-local-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
    NovaEnableRbdBackend: false
  ServicesDefault:
--- a/roles/ComputeOvsDpdk.yaml
+++ b/roles/ComputeOvsDpdk.yaml
@ -14,6 +14,7 @@
      subnet: storage_subnet
  deprecated_nic_config_name: 'compute-dpdk.yaml'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    VhostuserSocketGroup: "hugetlbfs"
    TunedProfileName: "cpu-partitioning"
    NovaLibvirtRxQueueSize: 1024
--- a/roles/ComputeOvsDpdkRT.yaml
+++ b/roles/ComputeOvsDpdkRT.yaml
@ -14,6 +14,7 @@
      subnet: storage_subnet
  ImageDefault: overcloud-realtime-compute
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    VhostuserSocketGroup: "hugetlbfs"
    TunedProfileName: "realtime-virtual-host"
    NovaLibvirtRxQueueSize: 1024
--- a/roles/ComputeOvsDpdkSriov.yaml
+++ b/roles/ComputeOvsDpdkSriov.yaml
@ -10,6 +10,7 @@
    - Tenant
    - Storage
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    VhostuserSocketGroup: "hugetlbfs"
    TunedProfileName: "cpu-partitioning"
    NovaLibvirtRxQueueSize: 1024
--- a/roles/ComputeOvsDpdkSriovRT.yaml
+++ b/roles/ComputeOvsDpdkSriovRT.yaml
@ -11,6 +11,7 @@
    - Storage
  ImageDefault: overcloud-realtime-compute
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    VhostuserSocketGroup: "hugetlbfs"
    TunedProfileName: "realtime-virtual-host"
    NovaLibvirtRxQueueSize: 1024
--- a/roles/ComputePPC64LE.yaml
+++ b/roles/ComputePPC64LE.yaml
@ -15,6 +15,7 @@
  HostnameFormatDefault: '%stackname%-novacomputeppc64le-%index%'
  ImageDefault: ppc64le-overcloud-full
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeRBDEphemeral.yaml
+++ b/roles/ComputeRBDEphemeral.yaml
@ -18,6 +18,7 @@
  update_serial: 25
  HostnameFormatDefault: '%stackname%-novacompute-rbd-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
    NovaEnableRbdBackend: true
  ServicesDefault:
--- a/roles/ComputeRealTime.yaml
+++ b/roles/ComputeRealTime.yaml
@ -19,6 +19,7 @@
  HostnameFormatDefault: '%stackname%-computerealtime-%index%'
  ImageDefault: overcloud-realtime-compute
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "realtime-virtual-host"
    KernelArgs: ""                  # these must be set in an environment file or
    IsolCpusList: ""                # similar according to the hardware of
--- a/roles/ComputeSriov.yaml
+++ b/roles/ComputeSriov.yaml
@ -13,6 +13,7 @@
    Storage:
      subnet: storage_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "cpu-partitioning"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeSriovIB.yaml
+++ b/roles/ComputeSriovIB.yaml
@ -13,6 +13,7 @@
    Storage:
      subnet: storage_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "cpu-partitioning"
  update_serial: 25
  ServicesDefault:
--- a/roles/ComputeSriovRT.yaml
+++ b/roles/ComputeSriovRT.yaml
@ -14,6 +14,7 @@
      subnet: storage_subnet
  ImageDefault: overcloud-realtime-compute
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "realtime-virtual-host"
    NovaLibvirtMemStatsPeriodSeconds: 0
    NovaComputeDisableIrqBalance: True
--- a/roles/DistributedCompute.yaml
+++ b/roles/DistributedCompute.yaml
@ -13,6 +13,7 @@
    Storage:
      subnet: storage_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles/DistributedComputeHCI.yaml
+++ b/roles/DistributedComputeHCI.yaml
@ -14,6 +14,7 @@
    StorageMgmt:
      subnet: storage_mgmt_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "throughput-performance"
  # CephOSD present so serial has to be 1
  update_serial: 1
--- a/roles/DistributedComputeHCIDashboard.yaml
+++ b/roles/DistributedComputeHCIDashboard.yaml
@ -18,6 +18,7 @@
    StorageDashboard:
      subnet: storage_dashboard_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "throughput-performance"
  # CephOSD present so serial has to be 1
  update_serial: 1
--- a/roles/DistributedComputeHCIScaleOut.yaml
+++ b/roles/DistributedComputeHCIScaleOut.yaml
@ -14,6 +14,7 @@
    StorageMgmt:
      subnet: storage_mgmt_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "throughput-performance"
  # CephOSD present so serial has to be 1
  update_serial: 1
--- a/roles/DistributedComputeScaleOut.yaml
+++ b/roles/DistributedComputeScaleOut.yaml
@ -13,6 +13,7 @@
    Storage:
      subnet: storage_subnet
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  update_serial: 25
  ServicesDefault:
--- a/roles_data.yaml
+++ b/roles_data.yaml
@ -213,6 +213,7 @@
      subnet: storage_subnet
  HostnameFormatDefault: '%stackname%-novacompute-%index%'
  RoleParametersDefault:
+    FsAioMaxNumber: 1048576
    TunedProfileName: "virtual-host"
  # Deprecated & backward-compatible values (FIXME: Make parameters consistent)
  # Set uses_deprecated_params to True if any deprecated params are used.