Merge pull request #227 from wayyoungboy/2.2.0-qulei

2.2.0 add rca scene and uodate check

cmd.py

@@ -627,6 +627,7 @@ def __init__(self):
         self.parser.add_option('--cases', type='string', help="check observer's cases on package_file")
         self.parser.add_option('--obproxy_cases', type='string', help="check obproxy's cases on package_file")
         self.parser.add_option('--store_dir', type='string', help='the dir to store check result, current dir by default.', default='./check_report/')
+        self.parser.add_option('--report_type', type='string', help='The type of the check report, support "table", "json", "xml", "yaml". default table', default='table')
         self.parser.add_option('-c', type='string', help='obdiag custom config', default=os.path.expanduser('~/.obdiag/config.yml'))
 
     def init(self, cmd, args):

docs/check.md

@@ -27,6 +27,17 @@ Example:
 obdiag check --cases= ad
 obdiag check --obproxy_cases= proxy
 obdiag check --cases=ad --obproxy_cases=proxy
+
+--report_type={ReportType}
+
+report_type是生成报告的格式，目前支持"table", "json", "xml"三种格式
+若未设定则默认为table
+
+--store_dir={StoreDir}
+
+store_dir是生成报告的存放路径，若未设定则默认为当前工作目录下的check_report文件夹内
+
+
 ```
 
 ### 关联持久化参数：

handler/checker/check_handler.py

@@ -133,6 +133,11 @@ def handle(self):
             if not os.path.exists(self.export_report_path):
                 self.stdio.warn("{0} not exists. mkdir it!".format(self.export_report_path))
                 os.mkdir(self.export_report_path)
+            # change self.export_report_type
+            if Util.get_option(self.options, 'report_type'):
+                self.export_report_type = Util.get_option(self.options, 'report_type')
+                if self.export_report_type not in ["table", "json", "xml", "yaml"]:
+                    raise CheckException("report_type must be table, json, xml, yaml")
             self.stdio.verbose("export_report_path is " + self.export_report_path)
 
             # get package's by package_name

handler/checker/check_report.py

@@ -50,7 +50,7 @@ def __init__(self, context, report_target="observer", export_report_path="./chec
         now = datetime.datetime.now()
         date_format = now.strftime("%Y-%m-%d-%H-%M-%S")
 
-        file_name = "/check_report_{0}_".format(report_target) + date_format
+        file_name = "/obdiag_check_report_{0}_".format(report_target) + date_format
         self.report_target = report_target
 
         report_path = self.export_report_path + file_name

handler/checker/step/sleep.py

@@ -0,0 +1,70 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*
+# Copyright (c) 2022 OceanBase
+# OceanBase Diagnostic Tool is licensed under Mulan PSL v2.
+# You can use this software according to the terms and conditions of the Mulan PSL v2.
+# You may obtain a copy of Mulan PSL v2 at:
+#          http://license.coscl.org.cn/MulanPSL2
+# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
+# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
+# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
+# See the Mulan PSL v2 for more details.
+
+"""
+@time: 2024/05/28
+@file: sleep.py
+@desc:
+"""
+import time
+
+from handler.checker.check_exception import StepExecuteFailException
+from common.ob_connector import OBConnector
+from common.tool import StringUtils
+from common.tool import Util
+
+
+class StepSleepHandler:
+    def __init__(self, context, step, task_variable_dict):
+        self.sleep_time = None
+        try:
+            self.context = context
+            self.stdio = context.stdio
+            self.ob_cluster = self.context.cluster_config
+            self.ob_cluster_name = self.ob_cluster.get("cluster_name")
+            self.tenant_mode = None
+            self.sys_database = None
+            self.database = None
+            self.ob_connector_pool = self.context.get_variable('check_obConnector_pool', None)
+            if self.ob_connector_pool is not None:
+                self.ob_connector = self.ob_connector_pool.get_connection()
+            if self.ob_connector is None:
+                raise Exception("self.ob_connector is None.")
+        except Exception as e:
+            self.stdio.error("StepSleepHandler init fail. Please check the OBCLUSTER conf. Exception : {0} .".format(e))
+            raise Exception("StepSleepHandler init fail. Please check the OBCLUSTER conf. Exception : {0} .".format(e))
+        self.task_variable_dict = task_variable_dict
+        self.enable_dump_db = False
+        self.trace_id = None
+        self.STAT_NAME = {}
+        self.report_file_path = ""
+        self.enable_fast_dump = False
+        self.ob_major_version = None
+        self.step = step
+
+    def execute(self):
+        try:
+            self.sleep_time = self.step.get("sleep", None)
+            if self.sleep_time == None:
+                raise StepExecuteFailException("StepSleepHandler execute sleep is not set")
+            if type(self.sleep_time) != int:
+                raise StepExecuteFailException("StepSleepHandler execute sleep type must be int")
+            time.sleep(int(self.sleep_time))
+            self.stdio.verbose("StepSleepHandler execute: {0}".format(self.sleep_time))
+        except Exception as e:
+            self.stdio.error("StepSleepHandler execute Exception: {0}".format(e))
+            raise StepExecuteFailException("StepSleepHandler execute Exception: {0}".format(e))
+        finally:
+            self.ob_connector_pool.release_connection(self.ob_connector)
+
+    def update_step_variable_dict(self):
+        return self.task_variable_dict

handler/checker/tasks/observer/clog/clog_disk_full.yaml

@@ -0,0 +1,10 @@
+info: "Check if there is a problem with clog disk full"
+task:
+  - version: "[4.0.0.0,*]"
+    steps:
+    - type: sql
+      sql: 'SELECT  GROUP_CONCAT(DISTINCT u.tenant_id) FROM oceanbase.gv$ob_units u  JOIN (    SELECT SVR_IP, SVR_PORT, TENANT_ID, value/100 AS value  FROM oceanbase.GV$OB_PARAMETERS   WHERE name = "log_disk_utilization_threshold") as c     ON u.SVR_IP = c.SVR_IP AND u.SVR_PORT = c.SVR_PORT AND u.TENANT_ID = c.TENANT_ID WHERE u.LOG_DISK_IN_USE > u.LOG_DISK_SIZE * c.value;'
+      result:
+        set_value: tenant_ids
+        verify: '[ -z "$tenant_ids" ]'
+        err_msg: "The following tenants have experienced clog disk full: #{tenant_ids}. Please check by obdiag rca --scene=clog_disk_full "

handler/checker/tasks/observer/column_storage/tenant_parameters.yaml

@@ -0,0 +1,11 @@
+info: 'Check whether there is any observer have CPU oversold.'
+task:
+  - version: "[4.3.1.0,*]"
+    steps:
+    - type: sql
+      sql: "SELECT GROUP_CONCAT(CONCAT(SVR_IP, ':', SVR_PORT) SEPARATOR ', ') AS IP_PORT_COMBINATIONSFROM from oceanbase.GV$OB_SERVERS WHERE CPU_ASSIGNED > CPU_CAPACITY;"
+      result:
+        set_value: CPU_oversold
+        verify: '[ -z "$CPU_oversold" ]'
+        report_type: warning
+        err_msg: 'Some observers have CPU oversold. There are #{CPU_oversold}'

handler/rca/plugins/gather.py

@@ -43,6 +43,7 @@ def init_parameters(self):
         self.conf_map["gather_scope"] = ""
         self.conf_map["store_dir"] = self.work_path
         self.conf_map["gather_target"] = "observer"
+        self.greps_key = []
 
     def grep(self, key):
         if key is None or len(key) < 1 or type(key) != str: