MetricsPrometheus::run() may panic

Bug Report

Please answer these questions before submitting your issue. Thanks!

The std::__1::__function::__func ... stuff is actually FunctionTimerTask::create([this] { run(); })

2023-10-29 15:54:56
[2023/10/29 07:54:56.876 +00:00] [ERROR] [BaseDaemon.cpp:422] ["Address not mapped to object."] [source=BaseDaemon] [thread_id=4761]
2023-10-29 15:54:56
[2023/10/29 07:54:56.876 +00:00] [ERROR] [BaseDaemon.cpp:407] ["Address: 0x90"] [source=BaseDaemon] [thread_id=4761]
2023-10-29 15:54:56
[2023/10/29 07:54:56.876 +00:00] [ERROR] [BaseDaemon.cpp:377] ["(from thread 4760) Received signal Segmentation fault(11)."] [source=BaseDaemon] [thread_id=4761]
2023-10-29 15:54:56
[2023/10/29 07:54:56.876 +00:00] [ERROR] [BaseDaemon.cpp:376] [########################################] [source=BaseDaemon] [thread_id=4761]
[2023/10/29 07:54:56.882 +00:00] [ERROR] [BaseDaemon.cpp:569] ["
       0x684f2e0    faultSignalHandler(int, siginfo_t*, void*) [tiflash+109376224]
                    libs/libdaemon/src/BaseDaemon.cpp:220
  0xffffb51bc81c    <unknown symbol> [linux-vdso.so.1+2076]
       0x19ad038    std::__1::__function::__func<DB::MetricsPrometheus::MetricsPrometheus(DB::Context&, DB::AsynchronousMetrics const&)::$_10, std::__1::allocator<DB::MetricsPrometheus::MetricsPrometheus(DB::Context&, DB::AsynchronousMetrics const&)::$_10>, void ()>::operator()() [tiflash+26923064]
                    /usr/local/bin/../include/c++/v1/__functional/function.h:345
       0x79d26f0    Poco::Util::TaskNotification::execute() [tiflash+127739632]
                    contrib/poco/Util/src/Timer.cpp:126
       0x79d2bb4    Poco::Util::FixedRateTaskNotification::execute() [tiflash+127740852]
                    contrib/poco/Util/src/Timer.cpp:199
       0x19ab040    DB::Timer::run() [tiflash+26914880]
                    dbms/src/Common/Timer.h:32
       0x7a5af04    Poco::ThreadImpl::runnableEntry(void*) [tiflash+128298756]
                    contrib/poco/Foundation/src/Thread_STD.cpp:139
       0x7a5bfd8    void* std::__1::__thread_proxy<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct> >, void* (*)(void*), Poco::ThreadImpl*> >(void*) [tiflash+128303064]
                    /usr/local/bin/../include/c++/v1/thread:291
  0xffffb112aa28    start_thread [libc.so.6+535080]"] [source=BaseDaemon] [thread_id=4761]

1. Minimal reproduce step (Required)

2. What did you expect to see? (Required)

3. What did you see instead (Required)

4. What is your TiFlash version? (Required)

Start at MetricsPrometheus::run() and the FunctionTimerTask::create([this] { run(); }) callback shown in the report. Read dbms/src/Common/Timer.h and the referenced timer and daemon stack frames, then determine the object-lifetime path that leads to the segmentation fault. Done means the failure is reproducible or explained and the scheduled metrics task no longer accesses invalid state.