diff --git a/server/agent_config/README-CH.md b/server/agent_config/README-CH.md index 4b59e41a024..d2cf68e43c4 100644 --- a/server/agent_config/README-CH.md +++ b/server/agent_config/README-CH.md @@ -2192,10 +2192,10 @@ inputs: **详细描述**: -当 deepflow-agent 在 Java 进程的函数调用栈中发现未能解析的函数名时，将触发进程函数符号表的再生成 -过程，而由于 Java 使用了 JIT 编译机制，符号表的再生成过程将延迟一定的时间。 - -TODO +当 deepflow-agent 在 Java 进程的函数调用栈中发现未能解析的函数名时，将触发进程函数符号表的生成和符号缓存 +的更新。当前 Java 符号文件是采用持续更新的方式，该 duration 用于控制推迟使用符号文件更新符号缓存的时间。 +原因是由于 Java 使用了 JIT 编译机制，编译符号生成有个预热阶段，为了获取更充足的 Java 符号需要推迟一段时间 +来更新 Java 符号的缓存，也可避免由于符号缺失而造成的频繁符号缓存刷新引起大量CPU资源消耗。 ##### 符号表文件最大大小 {#inputs.proc.symbol_table.java.max_symbol_file_size} @@ -3608,6 +3608,7 @@ inputs: **详细描述**: TCP 和 UDP 的端口白名单列表，白名单生效优先级低于 kprobe 黑名单。 +未列入黑名单、白名单的端口用 kprobe 做采集。 配置样例: `ports: 80,1000-2000` diff --git a/server/agent_config/README.md b/server/agent_config/README.md index 62ace189710..2bf34c9f5fc 100644 --- a/server/agent_config/README.md +++ b/server/agent_config/README.md @@ -2220,16 +2220,14 @@ inputs: **Description**: -When deepflow-agent finds that an unresolved function name appears in the function call -stack of a Java process, it will trigger the regeneration of the symbol file of the -process. Because Java utilizes the Just-In-Time (JIT) compilation mechanism, to obtain -more symbols for Java processes, the regeneration will be deferred for a period of time. - -At the startup of a Java program, the JVM and JIT compiler are in a "warm-up" phase. During this -period, symbol changes are typically frequent due to the dynamic compilation and optimization -processes. Therefore, deepflow-agent delay symbol collection for one minute after the Java program -starts, allowing the JVM and JIT to "warm up" and for symbol churn to be minimized before proceeding -with the collection. +When the `deepflow-agent` detects unresolved function names in the Java process call stack, it +triggers the generation of the process function symbol table and updates the symbol cache. Currently, +the Java symbol file is continuously updated, and the `duration` is used to control the delay in +updating the symbol cache with the new symbol file. This delay is necessary because Java uses a JIT +(Just-In-Time) compilation mechanism, which requires a warm-up phase for symbol generation. To obtain +more complete Java symbols, the update of the Java symbol cache is deferred. This approach also helps +avoid frequent symbol cache refreshes due to missing symbols, which could otherwise result in significant +CPU resource consumption. ##### Maximum Symbol File Size {#inputs.proc.symbol_table.java.max_symbol_file_size} @@ -3667,6 +3665,7 @@ inputs: **Description**: TCP&UDP Port Blacklist, Priority higher than kprobe-whitelist. +Use kprobe to collect data on ports that are not in the blacklist or whitelist. Example: `ports: 80,1000-2000` diff --git a/server/agent_config/template.yaml b/server/agent_config/template.yaml index 9259bf3f81a..72d534269d0 100644 --- a/server/agent_config/template.yaml +++ b/server/agent_config/template.yaml @@ -1325,22 +1325,21 @@ inputs: # ee_feature: false # description: # en: |- - # When deepflow-agent finds that an unresolved function name appears in the function call - # stack of a Java process, it will trigger the regeneration of the symbol file of the - # process. Because Java utilizes the Just-In-Time (JIT) compilation mechanism, to obtain - # more symbols for Java processes, the regeneration will be deferred for a period of time. - # - # At the startup of a Java program, the JVM and JIT compiler are in a "warm-up" phase. During this - # period, symbol changes are typically frequent due to the dynamic compilation and optimization - # processes. Therefore, deepflow-agent delay symbol collection for one minute after the Java program - # starts, allowing the JVM and JIT to "warm up" and for symbol churn to be minimized before proceeding - # with the collection. + # When the `deepflow-agent` detects unresolved function names in the Java process call stack, it + # triggers the generation of the process function symbol table and updates the symbol cache. Currently, + # the Java symbol file is continuously updated, and the `duration` is used to control the delay in + # updating the symbol cache with the new symbol file. This delay is necessary because Java uses a JIT + # (Just-In-Time) compilation mechanism, which requires a warm-up phase for symbol generation. To obtain + # more complete Java symbols, the update of the Java symbol cache is deferred. This approach also helps + # avoid frequent symbol cache refreshes due to missing symbols, which could otherwise result in significant + # CPU resource consumption. # ch: |- - # 当 deepflow-agent 在 Java 进程的函数调用栈中发现未能解析的函数名时，将触发进程函数符号表的再生成 - # 过程，而由于 Java 使用了 JIT 编译机制，符号表的再生成过程将延迟一定的时间。 + # 当 deepflow-agent 在 Java 进程的函数调用栈中发现未能解析的函数名时，将触发进程函数符号表的生成和符号缓存 + # 的更新。当前 Java 符号文件是采用持续更新的方式，该 duration 用于控制推迟使用符号文件更新符号缓存的时间。 + # 原因是由于 Java 使用了 JIT 编译机制，编译符号生成有个预热阶段，为了获取更充足的 Java 符号需要推迟一段时间 + # 来更新 Java 符号的缓存，也可避免由于符号缺失而造成的频繁符号缓存刷新引起大量CPU资源消耗。 # # upgrade_from: static_config.ebpf.java-symbol-file-refresh-defer-interval - # TODO: 待确认——该 duration 是否用于控制符号表再生成的延迟时间？ refresh_defer_duration: 60s # type: int # name: @@ -2358,14 +2357,15 @@ inputs: # description: # en: |- # TCP&UDP Port Whitelist, Priority lower than kprobe-blacklist. + # Use kprobe to collect data on ports that are not in the blacklist or whitelist. # # Example: `ports: 80,1000-2000` # ch: |- # TCP 和 UDP 的端口白名单列表，白名单生效优先级低于 kprobe 黑名单。 + # 未列入黑名单、白名单的端口用 kprobe 做采集。 # # 配置样例: `ports: 80,1000-2000` # upgrade_from: static_config.ebpf.kprobe-whitelist.port-list - # TODO: 未列入黑名单、白名单的端口是否用 kprobe 做采集？ ports: "" # type: section # name: