Berkeley DB | Forum profile

Hi, There's nothing serious about these warnings. Those uninitialized byes are for padding purposes. Please see the "--enable-umrw" build option for the Berkeley DB libraries: http://www.oracle.com/technology/documentation/berkeley-db/db/programmer_reference/build_unix_conf.html and use it when building the libraries to avoid seeing these errors. Regards, Andrei

Are you sure you are linking with the correct version of the library? That is something to check. Thanks, Sandra

This time i create a table Key : 5000 char , first 4000 chars are all character 'a', the left 1000 chars are generated randomly. Data: equal to key mean: db->put( dp, NULL, &key, &key, 0 ); still i didn't find any difference after i turn on the set_bt_compress funcion? Does anyone can help me, because now i am dealing a huge database (over 20 TB ), i need to compress it.

Hi, Andrei, I tried db-4.8.24, but problem still exists. Here are env files, we can still see the unrecycled extent file(__dbq.test1.13,__dbq.test1.14): # ls -lh /data1/testenv/ total 537M -rw-r----- 1 root root 24K Dec 10 12:51 __db.001 -rw-r----- 1 root root 35M Dec 10 12:51 __db.002 -rw-r----- 1 root root 321M Dec 10 12:51 __db.003 -rw-r----- 1 root root 4.2M Dec 10 12:51 __db.004 -rw-r----- 1 root root 35M Dec 10 12:51 __db.005 -rw-r-----...

Hi Sandra, Thank you very much for your answer. One would expect, that the process/thread is removed from the thread-environment, as soon as the process/thread disconnects. But it isn't. `db_stat -e -h /tmp/db | grep process/thread` shows also processes, which have disconnected long time ago. Greetings

I'm not sure I understand when exactly one of your application's threads blocks on a lock when trying to open a database. Is it right after the application restarts due to the failure or later on? It happens when restarting the application. The application opens the environment and hangs when opening the database. Have you run recovery after the failure? Here is info on handling failure in TDS apps:...

Of course, I forgot to mention that you can configure the cache-size before you open the database. The bigger the cache size, the better. However, you might not find a big help in configuring the cache, as you have a heavy number of inserts, but no updates and no deletes. One more thing, the data sizes you have may be large enough to warrant using Hash. When you get down to tuning of your application, you would want to bake-off Hash and Btree...

Hi Bogdan, What's the OS/File system you use? I'm using Windows 7 Professional 64 bits and the file system is NTFS in 4kb blocks. Is there any data access pattern? There will be no concurrent access, no transactions or threads. I'm expecting more than 10.000.000 blobs. Heavy number of inserts, no updates and no deletes. Thanks, Bruno

No, cache parameters are not set, they are default. So it is strange. Here is the code: env_flags = DB_CREATE | DB_INIT_TXN | DB_INIT_LOCK | DB_INIT_LOG | DB_INIT_MPOOL | DB_THREAD | DB_REGISTER | DB_RECOVER; rc = db_env_create(&dbenv, 0); rc = dbenv->set_lk_detect(dbenv, DB_LOCK_MINWRITE); dbenv->set_event_notify(dbenv, db_event_callback); rc = dbenv->open(dbenv, work_dir, env_flags, 0); error checks are omitted. It...

Hi, The reason we limit the maximum page size to 64kb is that we use unsigned 32-bit integers to indicate the offset in a page. If we increase the maximum page size to 128kb, many things need to change, and it may cause some potential risks. We need to evaluate how many advantages the changes will take. It does not just modify a constant simply.