Don’t use -dbcache for BDB anymore #2422

1. 
   sipa commented at 2:07 am on March 29, 2013: member

   -dbcache was originally used to set the maximum buffer size in the BDB environment, and was later changed to set the chainstate cache and leveldb caches. No need to use it for BDB now that only the wallet remains there.

   This should reduce memory allocation (but not necessarily memory usage) a bit, especially when -dbcache is set high.

2. Don't use -dbcache for BDB anymore

   -dbcache was originally used to set the maximum buffer size in the
   BDB environment, and was later changed to set the chainstate cache
   and leveldb caches. No need to use it for BDB now that only the
   wallet remains there.
   
   This should reduce memory allocation (but not necessarily memory
   usage) a bit.

   2e3c76bf98
3. 
   jgarzik commented at 2:16 am on March 29, 2013: contributor

   Does init.cpp want updating, too?
4. 
   sipa commented at 2:26 am on March 29, 2013: member

   @jgarzik Not that know of?
5. 
   gavinandresen commented at 6:51 pm on March 29, 2013: contributor

   ACK
6. 
   Diapolo commented at 8:42 pm on March 29, 2013: none

   Perhaps @jgarzik was talking about the help message? Set database cache size in megabytes (default: 25)
7. 
   sipa commented at 9:38 pm on March 29, 2013: member

   @Diapolo @jgarzik -dbcache is still the intended database cache, but we shouldn’t use that number both for the coincache/leveldb AND bdb.
8. in src/db.cpp: in 2e3c76bf98

   74@@ -75,11 +75,10 @@ bool CDBEnv::Open(const boost::filesystem::path& path)
   75     if (GetBoolArg("-privdb", true))
   76         nEnvFlags |= DB_PRIVATE;
   77 
   78-    unsigned int nDbCache = GetArg("-dbcache", 25);
   79     dbenv.set_lg_dir(pathLogDir.string().c_str());
   80-    dbenv.set_cachesize(nDbCache / 1024, (nDbCache % 1024)*1048576, 1);
   81-    dbenv.set_lg_bsize(1048576);
   82-    dbenv.set_lg_max(10485760);
   83+    dbenv.set_cachesize(0, 0x100000, 1); // 1 MiB should be enough for just the wallet
   

   ---

   ---

   

   rebroad commented at 2:39 am on March 30, 2013:

   When would 1MB not be sufficient?

   ---

   

   sipa commented at 4:52 pm on March 30, 2013:

   The wallet is always loaded entirely into RAM anyway, and only modified in small incremental ways, so a large cache shouldn’t ever be needed. However, when BDB was still used for the block chain database, we did require large caches to speed up prevout lookups during block validation. That part is now in LevelDB anyway.
9. 
   TheBlueMatt commented at 7:01 am on March 30, 2013: member

   ACK
10. 
    BitcoinPullTester commented at 8:32 am on March 30, 2013: none

    Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/2e3c76bf9857f5ac2814df9fd6b9b5b7a7eff88a for binaries and test log. This is an automated test script which runs test cases on each commit every time is updated. It, however, dies sometimes and fails to test properly, if you are waiting on a test, please check timestamps and if the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/ and contact BlueMatt on freenode if something looks broken.
11. 
    jgarzik commented at 4:17 am on March 31, 2013: contributor

    ACK
12. sipa referenced this in commit 767fe14a27 on Mar 31, 2013
13. sipa merged this on Mar 31, 2013
14. sipa closed this on Mar 31, 2013

    ---

15. sipa deleted the branch on May 3, 2013
16. laudney referenced this in commit 0e774a6068 on Mar 19, 2014
17. DrahtBot locked this on Sep 8, 2021

Contributors
sipa jgarzik gavinandresen Diapolo rebroad TheBlueMatt BitcoinPullTester