[Bug] meta_tool will coredump when used to handle tablet meta

[wenzhenghu]

Search before asking

• I had searched in the issues and found no similar issues.

Version

3.0 and higher

What's Wrong?

           PID: 2248042 (meta_tool)
           UID: 0 (root)
           GID: 0 (root)
        Signal: 11 (SEGV)
     Timestamp: Tue 2026-03-17 17:27:37 CST (15h ago)
  Command Line: /mnt/wzh-hdd/hy-doris/be/output/lib/meta_tool --operation=show_meta --pb_meta_path=1773380615252.hdr
    Executable: /mnt/wzh-hdd/hy-doris/be/output/lib/meta_tool
 Control Group: /user.slice/user-1002.slice/session-141.scope
          Unit: session-141.scope
         Slice: user-1002.slice
       Session: 141
     Owner UID: 1002 (wzh)
       Boot ID: 2e19417b314940549429216848c47e72
    Machine ID: 2d615b97f292443ebe9dfb7dc34eb547
      Hostname: hydoris-compute
       Storage: /var/lib/systemd/coredump/core.meta_tool.0.2e19417b314940549429216848c47e72.2248042.1773739657000000.zst (present)
  Size on Disk: 3.0M
       Message: Process 2248042 (meta_tool) of user 0 dumped core.

                Module libgcc_s.so.1 from deb gcc-14-14.2.0-4ubuntu2~24.04.1.amd64
                Module libstdc++.so.6 from deb gcc-14-14.2.0-4ubuntu2~24.04.1.amd64
                Module meta_tool without build-id.
                Stack trace of thread 2248042:
                #0  0x00006317c6925d89 _ZN5doris17TabletSchemaCache6insertERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE (meta_tool + 0x11315d89)
                #1  0x00006317c68c988f _ZN5doris10TabletMeta12init_from_pbERKNS_12TabletMetaPBE (meta_tool + 0x112b988f)
                #2  0x00006317c68cb533 _ZN5doris10TabletMeta16create_from_fileERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE (meta_tool + 0x112bb533)
                #3  0x00006317c1328462 _Z9show_metav (meta_tool + 0xbd18462)
                #4  0x00006317c132eea7 main (meta_tool + 0xbd1eea7)
                #5  0x00007b1cae42a1ca __libc_start_call_main (libc.so.6 + 0x2a1ca)
                #6  0x00007b1cae42a28b __libc_start_main_impl (libc.so.6 + 0x2a28b)
                #7  0x00006317c132802a _start (meta_tool + 0xbd1802a)

                Stack trace of thread 2248043:
                #0  0x00007b1cae4ecadf __GI___clock_nanosleep (libc.so.6 + 0xecadf)
                #1  0x00007b1cae4f9a27 __GI___nanosleep (libc.so.6 + 0xf9a27)
                #2  0x00007b1cae52975c usleep (libc.so.6 + 0x12975c)
                #3  0x00006317cb13a589 _ZN4bvar6detail16SamplerCollector3runEv (meta_tool + 0x15b2a589)
                #4  0x00006317cb13c519 _ZN4bvar6detail16SamplerCollector15sampling_threadEPv (meta_tool + 0x15b2c519)
                #5  0x00007b1cae49caa4 start_thread (libc.so.6 + 0x9caa4)
                #6  0x00007b1cae529c6c __clone3 (libc.so.6 + 0x129c6c)
                ELF object binary architecture: AMD x86-64

What You Expected?

do the right thing

How to Reproduce?

just use it

Anything Else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[Arun12-sketch]

Hi @wenzhenghu,
I'd like to work on this issue. Can you please assign it to me?
I've analyzed the coredump stack trace and identified the root cause:
The TabletSchemaCache singleton is never initialized in meta_tool's main().
When TabletMeta::init_from_pb() calls TabletSchemaCache::instance()->insert(),
it dereferences a nullptr, causing SIGSEGV.