Cut down mark_cache default size and add the corresponding metrics in…

… grafana (#7005) ref #6426
pingcap · Mar 13, 2023 · 3737e0c · 3737e0c
1 parent a7642ef
commit 3737e0c
Show file tree

Hide file tree

Showing 8 changed files with 164 additions and 13 deletions.
diff --git a/dbms/src/Core/Defines.h b/dbms/src/Core/Defines.h
@@ -107,7 +107,7 @@ static constexpr UInt64 DEFAULT_DISAGG_TASK_TIMEOUT_SEC = 3600;
 
 #define PLATFORM_NOT_SUPPORTED "The only supported platforms are x86_64 and AArch64 (work in progress)"
 
-#define DEFAULT_MARK_CACHE_SIZE (5ULL * 1024 * 1024 * 1024)
+#define DEFAULT_MARK_CACHE_SIZE (1ULL * 1024 * 1024 * 1024)
 
 #define DEFAULT_METRICS_PORT 8234
 

diff --git a/etc/config-template.toml b/etc/config-template.toml
@@ -19,9 +19,9 @@
 ## The TiFlash HTTP service port.
 # http_port = 8123
 ## The cache size limit of the metadata of a data block. Generally, you do not need to change this value.
-# mark_cache_size = 5368709120
+# mark_cache_size = 1073741824
 ## The cache size limit of the min-max index of a data block. Generally, you do not need to change this value.
-# minmax_index_cache_size = 5368709120
+# minmax_index_cache_size = 1073741824
 ## The path in which the TiFlash temporary files are stored. By default it is the first directory in storage.latest.dir appended with "/tmp".
 # tmp_path = "/tidb-data/tiflash-9000/tmp"
 

diff --git a/metrics/grafana/tiflash_summary.json b/metrics/grafana/tiflash_summary.json
@@ -7416,6 +7416,157 @@
           "percentage": false,
           "steppedLine": false,
           "description": "Information about schema of column file, to learn the memory usage of schema"
+        },
+        {
+          "aliasColors": {},
+          "dashLength": 10,
+          "datasource": "${DS_TEST-CLUSTER}",
+          "description": "cache misses or cache hits of mark_cache.\nBased on this infactor, we can check whether mark_cache is large enough",
+          "fieldConfig": {
+            "defaults": {},
+            "overrides": []
+          },
+          "fill": 1,
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 72
+          },
+          "id": 169,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "nullPointMode": "null",
+          "options": {
+            "alertThreshold": true
+          },
+          "pluginVersion": "7.5.11",
+          "pointradius": 2,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "targets": [
+            {
+              "exemplar": true,
+              "expr": "max(tiflash_system_profile_event_MarkCacheMisses{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\", instance=~\"$instance\"}) by (instance)",
+              "interval": "",
+              "legendFormat": "mark cache misses",
+              "queryType": "randomWalk",
+              "refId": "A"
+            },
+            {
+              "exemplar": true,
+              "expr": "max(tiflash_system_profile_event_MarkCacheHits{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\", instance=~\"$instance\"}) by (instance)",
+              "hide": false,
+              "interval": "",
+              "legendFormat": "mark cache hits",
+              "refId": "B"
+            }
+          ],
+          "thresholds": [],
+          "timeRegions": [],
+          "title": "Effectiveness of Mark Cache",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true,
+              "$$hashKey": "object:259"
+            },
+            {
+              "format": "percent",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false,
+              "$$hashKey": "object:260"
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          },
+          "bars": false,
+          "dashes": false,
+          "fillGradient": 0,
+          "hiddenSeries": false,
+          "percentage": false,
+          "points": false,
+          "stack": false,
+          "steppedLine": false,
+          "timeFrom": null,
+          "timeShift": null,
+          "transformations": [
+            {
+              "id": "calculateField",
+              "options": {
+                "mode": "binary",
+                "reduce": {
+                  "reducer": "sum"
+                },
+                "binary": {
+                  "left": "mark cache misses",
+                  "operator": "+",
+                  "right": "mark cache hits",
+                  "reducer": "sum"
+                },
+                "alias": "mark cache count total"
+              }
+            },
+            {
+              "id": "calculateField",
+              "options": {
+                "mode": "binary",
+                "reduce": {
+                  "reducer": "sum"
+                },
+                "binary": {
+                  "left": "mark cache hits",
+                  "operator": "/",
+                  "right": "mark cache count total",
+                  "reducer": "sum"
+                },
+                "alias": "mark cache effectiveness"
+              }
+            },
+            {
+              "id": "filterFieldsByName",
+              "options": {
+                "include": {
+                  "names": [
+                    "Time",
+                    "mark cache effectiveness"
+                  ]
+                }
+              }
+            }
+          ]
         }
       ],
       "repeat": null,

diff --git a/tests/docker/config/tics_dt.toml b/tests/docker/config/tics_dt.toml
@@ -15,8 +15,8 @@
 tmp_path = "/tmp/tiflash/data/tmp"
 path = "/tmp/tiflash/data/db"
 capacity = "107374182400" # 100GB
-mark_cache_size = 5368709120
-minmax_index_cache_size = 5368709120
+mark_cache_size = 1073741824
+minmax_index_cache_size = 1073741824
 tcp_port = 9000
 http_port = 8123
 

diff --git a/tests/docker/config/tiflash_dt.toml b/tests/docker/config/tiflash_dt.toml
@@ -22,8 +22,8 @@ capacity = "10737418240"
 # path = "/tmp/tiflash/data/db,/tmp/tiflash1,/tmp/tiflash2"
 # capacity = "0"
 
-mark_cache_size = 5368709120
-minmax_index_cache_size = 5368709120
+mark_cache_size = 1073741824
+minmax_index_cache_size = 1073741824
 tcp_port = 9000
 http_port = 8123
 

diff --git a/tests/docker/config/tiflash_dt_async_grpc.toml b/tests/docker/config/tiflash_dt_async_grpc.toml
@@ -17,8 +17,8 @@ tmp_path = "/tmp/tiflash/data/tmp"
 path = "/tmp/tiflash/data/db"
 capacity = "10737418240"
 
-mark_cache_size = 5368709120
-minmax_index_cache_size = 5368709120
+mark_cache_size = 1073741824
+minmax_index_cache_size = 1073741824
 tcp_port = 9000
 http_port = 8123
 

diff --git a/tests/docker/config/tiflash_dt_disable_local_tunnel.toml b/tests/docker/config/tiflash_dt_disable_local_tunnel.toml
@@ -17,8 +17,8 @@ tmp_path = "/tmp/tiflash/data/tmp"
 path = "/tmp/tiflash/data/db"
 capacity = "10737418240"
 
-mark_cache_size = 5368709120
-minmax_index_cache_size = 5368709120
+mark_cache_size = 1073741824
+minmax_index_cache_size = 1073741824
 tcp_port = 9000
 http_port = 8123
 

diff --git a/tests/docker/config/tiflash_dt_disable_planner.toml b/tests/docker/config/tiflash_dt_disable_planner.toml
@@ -17,8 +17,8 @@ tmp_path = "/tmp/tiflash/data/tmp"
 path = "/tmp/tiflash/data/db"
 capacity = "10737418240"
 
-mark_cache_size = 5368709120
-minmax_index_cache_size = 5368709120
+mark_cache_size = 1073741824
+minmax_index_cache_size = 1073741824
 tcp_port = 9000
 http_port = 8123