Multiprocess access

[chuckb]

I have one process that needs to read data only and one process that will be r/w to the same database. This is on a local Linux file system. I seem to be having some process instability doing this. Are there C flags that I should pass to my applications to insure this runs stably? Under what conditions may the reader encounter lock conditions?

[chuckb]

More specifically, if my read/write process is by itself, it works fine. Once I start up the read only process (it opens the db with the UNQLITE_OPEN_READONLY flag), after a short while, on a commit, I will get an error -14 (UNQLITE_BUSY). If I then retry the commit, I get a segfault.

[chuckb]

I was able to get a backtrace of the faulting r/w process. This fault occurs after a commit returns a -14, and then this fault occurs on the retried commit.

* thread #1, name = 'epoch-block-mat', stop reason = signal SIGSEGV: invalid address (fault address: 0x0)
    frame #0: 0x00005555555c1fe5 epoch-block-matcher`unqliteOsWrite(id=0x0000000000000000, pBuf=0x00007fffffffdb94, amt=4, offset=8) at unqlite.c:52480
   52477	}
   52478	UNQLITE_PRIVATE int unqliteOsWrite(unqlite_file *id, const void *pBuf, unqlite_int64 amt, unqlite_int64 offset)
   52479	{
-> 52480	  return id->pMethods->xWrite(id, pBuf, amt, offset);
   52481	}
   52482	UNQLITE_PRIVATE int unqliteOsTruncate(unqlite_file *id, unqlite_int64 size)
   52483	{
(lldb) bt
* thread #1, name = 'epoch-block-mat', stop reason = signal SIGSEGV: invalid address (fault address: 0x0)
  * frame #0: 0x00005555555c1fe5 epoch-block-matcher`unqliteOsWrite(id=0x0000000000000000, pBuf=0x00007fffffffdb94, amt=4, offset=8) at unqlite.c:52480
    frame #1: 0x00005555555c4198 epoch-block-matcher`WriteInt32(pFd=0x0000000000000000, iNum=4, iOfft=8) at unqlite.c:55574
    frame #2: 0x00005555555c6ce1 epoch-block-matcher`unqliteFinalizeJournal(pPager=0x0000555555a3e664, pRetry=0x00007fffffffdc00, close_jrnl=1) at unqlite.c:56961
    frame #3: 0x00005555555c72dd epoch-block-matcher`pager_commit_phase1(pPager=0x0000555555a3e664) at unqlite.c:57152
    frame #4: 0x00005555555c76c8 epoch-block-matcher`unqlitePagerCommit(pPager=0x0000555555a3e664) at unqlite.c:57290
    frame #5: 0x0000555555570a3f epoch-block-matcher`unqlite_commit(pDb=0x0000555555a406fc) at unqlite.c:6190
    frame #6: 0x00005555556857c9 epoch-block-matcher`KVPStore<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >::commit(this=0x0000555555a32fc0) at KVPStore.cpp:88
    frame #7: 0x00005555556ffc94 epoch-block-matcher`PriceEpochBlockMatcherDAO::commit(this=0x0000555555a3cc80) at PriceEpochBlockMatcherDAO.cpp:85
    frame #8: 0x00005555555cdb55 epoch-block-matcher`main(argc=1, argv=0x00007fffffffe4b8) at epoch-block-matcher.cpp:136
    frame #9: 0x00007ffff6aebc87 libc.so.6`__libc_start_main + 231
    frame #10: 0x000055555556d10a epoch-block-matcher`_start + 42

[chuckb]

If I do a rollback when I get the commit failure, then upon retry, I do not get a crash, but I get a repeated server busy error.

[symisc]

Hi Chuck,

This issue has been addressed on our latest bug fix. Refer to the issue: #137 for additional information. Please update to the latest patch fix.