Sensible max hsm concurrency values

vinil-v · Sep 7, 2021 · 636a636 · 636a636
1 parent a1c4e9a
commit 636a636
Show file tree

Hide file tree

Showing 7 changed files with 19 additions and 11 deletions.
diff --git a/blocks/lustre-cluster.json b/blocks/lustre-cluster.json
@@ -18,6 +18,7 @@
     "lustre_instances": 2,
     "lustre_storage_account": "<NOT-SET>",
     "lustre_storage_container": "hsm",
+    "lustre_storage_sas": "saskey.{{variables.lustre_storage_account}}.{{variables.lustre_storage_container}},rwdl,1y",
     "lustre_version": "<NOT-SET>",
     "lustre_mount": "/lustre",
     "lustre_avset": "{{variables.resource_group}}avset",
@@ -93,7 +94,8 @@
       "script": "lfsmaster.sh",
       "tag": "lustremaster",
       "args": [
-        "/dev/sdb"
+        "/dev/sdb",
+        "variables.lustre_instances"
       ],
       "sudo": true
     },
@@ -111,9 +113,8 @@
       "args": [
         "lustremaster",
         "variables.lustre_storage_account",
-        "sakey.{{variables.lustre_storage_account}}",
-        "variables.lustre_storage_container",
-        "variables.lustre_version"
+        "variables.lustre_storage_sas",
+        "variables.lustre_storage_container"
       ],
       "tag": "lustre",
       "sudo": true

diff --git a/examples/lustre_avset/config.json b/examples/lustre_avset/config.json
@@ -164,7 +164,8 @@
             "script": "lfsmaster.sh",
             "tag": "lustremaster",
             "args": [
-                "/dev/sdb"
+                "/dev/sdb",
+                "variables.lustre_instances"
             ],
             "sudo": true
         },

diff --git a/examples/lustre_combined/config.json b/examples/lustre_combined/config.json
@@ -145,7 +145,8 @@
             "script": "lfsmaster.sh",
             "tag": "lustre",
             "args": [
-                "/dev/sdb"
+                "/dev/sdb",
+                "variables.lustre_instances"
             ],
             "sudo": true
         },

diff --git a/examples/lustre_full/config.json b/examples/lustre_full/config.json
@@ -176,7 +176,8 @@
             "script": "lfsmaster.sh",
             "tag": "lfsmaster",
             "args": [
-                "/dev/nvme0n1"
+                "/dev/nvme0n1",
+                "variables.oss_instances"
             ],
             "sudo": true
         },

diff --git a/examples/lustre_ltop/config.json b/examples/lustre_ltop/config.json
@@ -137,7 +137,8 @@
             "script": "lfsmaster.sh",
             "tag": "lustre",
             "args": [
-                "/dev/sdb"
+                "/dev/sdb",
+                "variables.lustre_instances"
             ],
             "sudo": true
         },

diff --git a/examples/lustre_manageddisk/config.json b/examples/lustre_manageddisk/config.json
@@ -189,7 +189,8 @@
       "script": "lfsmaster.sh",
       "tag": "lfsmaster",
       "args": [
-        "/dev/sdc"
+        "/dev/sdc",
+        "variables.oss_instances"
       ],
       "sudo": true
     },

diff --git a/scripts/lfsmaster.sh b/scripts/lfsmaster.sh
@@ -2,7 +2,9 @@
 
 # arg: $1 = device (e.g. L=/dev/sdb Lv2=/dev/nvme0n1)
 device=$1
-hsm_max_requests=${2:-128}
+num_oss=$2
+# default of 3 concurrent hsm requests per oss
+reqs_per_oss=${3:-3}
 
 # this will only install MDS on first node in a scaleset
 if [ "$PSSH_NODENUM" = "0" ]; then
@@ -16,7 +18,7 @@ if [ "$PSSH_NODENUM" = "0" ]; then
     # set up hsm
     lctl set_param -P mdt.*-MDT0000.hsm_control=enabled
     lctl set_param -P mdt.*-MDT0000.hsm.default_archive_id=1
-    lctl set_param mdt.*-MDT0000.hsm.max_requests=$hsm_max_requests
+    lctl set_param mdt.*-MDT0000.hsm.max_requests=$(($num_oss * $reqs_per_oss))
 
     # allow any user and group ids to write
     lctl set_param mdt.*-MDT0000.identity_upcall=NONE