Hi Martin,

Thanks for getting back to me! Sorry I missed the information. Here goes:

OS is a Debian Wheeze, stable, here's the uname: 3.2.0-4-amd64 #1 SMP Debian 3.2.41-2+deb7u2 x86_64 GNU/Linux

The machine has 24 cores and 140GB RAM. There is space left on the partition, although I am suspicious if that was also true when it crashed the first time.

Server:

MonetDB 5 server v11.15.11 "Feb2013-SP3" (64-bit, 64-bit oids) Copyright (c) 1993-July 2008 CWI Copyright (c) August 2008-2013 MonetDB B.V., all rights reserved Visit http://www.monetdb.org/ for further information Found 142.0GiB available memory, 24 available cpu cores Libraries: libpcre: 8.30 2012-02-04 (compiled with 8.02) openssl: OpenSSL 0.9.8o 01 Jun 2010 (compiled with OpenSSL 0.9.8o 01 Jun 2010) libxml2: 2.7.8 (compiled with 2.7.8) Compiled by: root@dev.monetdb.org (x86_64-pc-linux-gnu) Compilation: gcc -O3 -fomit-frame-pointer -pipe -Wp,-D_FORTIFY_SOURCE=2 Linking : /usr/bin/ld -m elf_x86_64

I can tell you the schema is (from memory):

id int ip_addr inet ip_str varchar ip_start int ip_end int ptr varchar sld varchar tld varchar country varchar city varchar

I have made a very short input file available on

http://pkidata.net.in.tum.de/export.csv.tgz (33M)

BTW, there was a typo in my earlier mail: we have 133 billion, not million, rows.

Ralph

On 08/28/2013 09:26 PM, Martin Kersten wrote:

Hi

Please provide some background information, such as

• what operating system are you using
• what version of MonetDB
• if the input file is public, a reference to it, and the schema used.

regards, Martin

On 8/28/13 8:35 PM, Ralph Holz wrote:

...

Hi,

I've tried to load a good chunk of the Internet census into our MonetDB

• about 100M rows. Although the import scripts completed, it seems it

shot down Monet:

2013-08-28 04:10:35 MSG merovingian[9857]: database 'census' (10773) was killed by signal SIGBUS 2013-08-28 10:47:59 MSG merovingian[9857]: database 'census' has crashed after start on 2013-08-28 01:14:07, attempting restart, up min/avg/max: 4m/4d/3w, crash average: 1.00 0.10 0.03 (8-7=1)

This is then followed by

2013-08-28 10:48:32 MSG census[19161]: # still reading write-ahead log "/mnt/monetdb/census/sql_logs/sql/log.990" (0% done) [...] 2013-08-28 10:53:12 MSG census[19161]: # still reading write-ahead log "/mnt/monetdb/census/sql_logs/sql/log.990" (4% done) 2013-08-28 10:53:34 MSG merovingian[9857]: database 'census' (19161) has crashed (dumped core)

Restarting it, yielded the same results for a few times. Once or twice it got to 35% in the WAL

2013-08-28 15:34:27 MSG census[14420]: # still reading write-ahead log "/mnt/monetdb/census/sql_logs/sql/log.990" (35% done) 2013-08-28 15:34:59 MSG merovingian[9857]: database 'census' (14420) has crashed (dumped core) 2013-08-28 15:34:59 ERR merovingian[9857]: client error: database 'census' has crashed after starting, manual intervention needed, check monetdbd's logfile for details 2013-08-28 15:34:59 ERR merovingian[9857]: client error: client (local) sent challenge in incomplete block: 2013-08-28 15:34:59 ERR merovingian[9857]: client error: client (local) sent challenge in incomplete block: 2013-08-28 15:34:59 MSG merovingian[9857]: database 'census' has crashed after start on 2013-08-28 14:43:34, attempting restart, up min/avg/max: 4m/4d/3w, crash average: 1.00 0.90 0.30 (16-7=9)

Since then, it won't start (I don't know what caused the client error BTW).

Data loss is not a concern, but we'd sure like to know if this can be fixed or if we should start looking at another DB...

Thanks for any help, Ralph

---

users-list mailing list users-list@monetdb.org http://mail.monetdb.org/mailman/listinfo/users-list

Hi

Thanks, this little file contains 300.000 records. So your input file must be something like 433 GB I assume that you loaded the file with LOCKED mode, then at least you have a chance without the temporary copies needed.

With 9 fields of 8 bytes each, your storage footprint becomes easily 8 * 133 GB = 1.2TB, excluding the string value dictionaries. Strings may use less then 8 bytes each when there are a lot of duplicates.

To be on the safe size, I would aim for free disk-space of about 3-4TB

You might consider having a look at http://www.monetdb.org/Documentation/Cookbooks/SQLrecipes/storage-model to get a better estimate after loading your small sample.

On 8/28/13 10:48 PM, Ralph Holz wrote:

Hi,

...

I have made a very short input file available on

http://pkidata.net.in.tum.de/export.csv.tgz (33M)

The delimiter is #.

Looking at that file again, I must have missed some fields in the schema in my last mail. I think you can ignore those, they're mostly empty.

COPY into does not ignore them, they have to be specified. There is not (yet) a partial COPY into provided.

I have no clue what will happen if your scheme does not match the file, at least a good test.

regards, Martin

Ralph