Maxim Grinev | 6 Jul 11:43 2007
Picon

[ANN] SDBAdmin by Tony Scott

Hello!

We are pleased to announce a new admin GUI tool. It is a contribution by
Tony Scott (tony@...). Thanks a lot to Tony for his valuable
contribution!

SDBAdmin is a Windows-based GUI front-end for local or remote Sedna 
databases, written in C++.  It provides a simple interface to edit the Sedna
configuration, start and stop Sedna, create or drop databases,
collections, and indexes, and submit XQueries for evaluation. 
Documents and schema can be viewed by means of a simple point-and-
click, and queries can be saved for re-use.
Unique to SDBAdmin is a load/unload facility (in addition to support
for Sedna's import/export), which supports bulk-load of multiple
documents to/from a folder. Future releases will include support for
modules, users and roles. SDBAdmin is free software.

The tool is available from the Sedna download page: 
http://www.modis.ispras.ru/register.htm

Salute,
Maxim Grinev
Sedna team

Ivan Shcheklein | 10 Jul 15:01 2007
Picon

[ANN] Sedna 2.1 is available for download.

We are pleased to announce a version 2.1 of Sedna!

Changes:

  • Further indexes optimization;
  • Further bulk load optimization;
  • A lot of bug fixes and small improvments.

Note! Use export utility (see Sedna Administration Guide, section 2.5) to convert Sedna's 2.0 (or earlier) databases into Sedna 2.1 format.


Spencer Cheng | 12 Jul 18:18 2007

SE3046 session numbers exceed permissible threshold

I am running in this error when I try to parallelize the processing  
of multiple RSS feeds. Basically I am running a thread per feed using  
multiple instances of se_term when I run into this. The number of  
feeds I am experimenting with right now is about 20 but could grow  
significantly.

Is there a good way to avoid this problem (other than restricting the  
number of sessions by design)? I noticed that there is a parameter  
regarding the number of micro transactions but the documentation says  
"don't touch" :-)

Sedna version is 2.0.54 running uBuntu Edgy.

Regards,
Spencer

PS. I also get a bunch of SE4703.

Spencer Cheng | 12 Jul 23:11 2007

Re: SE3046 session numbers exceed permissible threshold

Some further notes. I can make both error (SE3046 and SE4703)  
disappear by restricting the number of sessions to be no  more than 3.

Am I coming across some limitations in current locking strategy?  
There is one collection which I am adding documents to from different  
threads but other than that I don't think I have any document and/or  
collection collisions.

Regards,
Spencer

On Jul 12, 2007, at 12:18, Spencer Cheng wrote:

> I am running in this error when I try to parallelize the processing
> of multiple RSS feeds. Basically I am running a thread per feed using
> multiple instances of se_term when I run into this. The number of
> feeds I am experimenting with right now is about 20 but could grow
> significantly.
>
> Is there a good way to avoid this problem (other than restricting the
> number of sessions by design)? I noticed that there is a parameter
> regarding the number of micro transactions but the documentation says
> "don't touch" :-)
>
> Sedna version is 2.0.54 running uBuntu Edgy.
>
> Regards,
> Spencer
>
>
> PS. I also get a bunch of SE4703.
>

Spencer Cheng | 14 Jul 02:34 2007

restore fails in version 2.1.2

I exported my test DB using the latest version (2.1.2) and try to  
restore it to a different DB. The restore failed and caused the SM  
and GOV to crash. I think my original DB is also corrupted as I can't  
restart it either.

Regards,
Spencer

-------------------------------
spencer <at> bayu:~/tmp$ se_exp restore test1 test-2007.07.13/

RESTORING DATA (path=test-2007.07.13/ host=localhost database=test1)

ERROR: failed to execute query:

declare boundary-space preserve; LOAD "57.xml" "b6aef090-13a5-012a- 
e7a6-0018f3cf4296-articles"

Details:

SEDNA Message: ERROR SE3007
Failed to recieve a message.
Details: Connection was broken while obtaining bulk load result

Gregg Reynolds | 14 Jul 20:28 2007

sedna ports?

Hi,

I'm new to sedna and would like to run it on OpenBSD and maybe OSX.
Is anybody doing any ports work?  The shared memory stuff would need
to be reworked to support SYSV IPC, and it wouldn't hurt to tweak some
of the build/config stuff to improve portability.  There may be other
things but I've only just started looking at the code.

Thanks,

gregg

Maria Rekouts | 15 Jul 08:57 2007
Picon

Re: restore fails in version 2.1.2

Spencer,

sorry, but it's difficult to say something without
reproducing the problem.

Can you give us this exported directory "test-2007.07.13"?
If it is not possible then maybe only "57.xml" would be
enough to reproduce the failure.

Thank you for this report,

Maria Rekouts
Sedna team

On Fri, 13 Jul 2007 20:34:10 -0400
 Spencer Cheng <scheng@...> wrote:
> I exported my test DB using the latest version (2.1.2)
> and try to  
> restore it to a different DB. The restore failed and
> caused the SM  
> and GOV to crash. I think my original DB is also
> corrupted as I can't  
> restart it either.
> 
> Regards,
> Spencer
> 
> -------------------------------
> spencer <at> bayu:~/tmp$ se_exp restore test1 test-2007.07.13/
> 
> RESTORING DATA (path=test-2007.07.13/ host=localhost
> database=test1)
> 
> ERROR: failed to execute query:
> 
> declare boundary-space preserve; LOAD "57.xml"
> "b6aef090-13a5-012a- 
> e7a6-0018f3cf4296-articles"
> 
> Details:
> 
> SEDNA Message: ERROR SE3007
> Failed to recieve a message.
> Details: Connection was broken while obtaining bulk load
> result
> 
> 
> 
>
-------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express
> and take
> control of your XML. No limits. Just data. Click to get
> it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Sedna-discussion mailing list
> Sedna-discussion@...
>
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

ZNV | 15 Jul 22:44 2007
Picon

Re: sedna ports?

Hello Gregg!
It seems to me that OBSD has mmap sys call, doesn't it? Anyway the problem is more severe than simply switching to another IPC API. OBSD does address space layout randomisation. Sedna assumes she is able to map the shared memory segment at the same address in all transaction processes. Even worse, she assumes this address won't change from run to run. We need to redo many code to fix this. The best thing for now is a quick and dirty workaround if one exists, since the true solution is complex. Beeng able to map at least 10 Mb at the constant address (determined during installation for instance) will be enough, any ideas how to achieve this?
And we still have no config stuff, shame on us :( The currently established solution works fine for Linux and Windows, however things must change if we are going to support more OSes officially. (I am advocating for CMake.) Things are done but sometimes very slow.
Thank you for your interest in Sedna, we do really need your feedback.

WBR
Mejedi, Sedna Team

2007/7/14, Gregg Reynolds <dev-5D4HdanPyQpiLUuM0BA3LQ@public.gmane.org >:
Hi,

I'm new to sedna and would like to run it on OpenBSD and maybe OSX.
Is anybody doing any ports work?  The shared memory stuff would need
to be reworked to support SYSV IPC, and it wouldn't hurt to tweak some
of the build/config stuff to improve portability.  There may be other
things but I've only just started looking at the code.

Thanks,

gregg

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Sedna-discussion mailing list
Sedna-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

Spencer Cheng | 15 Jul 23:06 2007

Re: restore fails in version 2.1.2

Maria,

In addition to the logs and backup I sent you, I also re-ran the  
offending transactions (according to the import log) interactively.  
The error is as follows.

spencer <at> bayu:~/tmp/test-2007.07.13$ se_term test
Welcome to term, the SEDNA Interactive Terminal. Type \? for help.
 > declare boundary-space preserve; LOAD "57.xml" "b6aef090-13a5-012a- 
e7a6-0018f3cf4296-articles"   &
SEDNA Message: ERROR SE1035
Memory mapping error.
Position: [vmm.cpp:_vmm_remap:280]
test>

Regards,
Spencer

On Jul 15, 2007, at 2:57, Maria Rekouts wrote:

> Spencer,
>
> sorry, but it's difficult to say something without
> reproducing the problem.
>
> Can you give us this exported directory "test-2007.07.13"?
> If it is not possible then maybe only "57.xml" would be
> enough to reproduce the failure.
>
> Thank you for this report,

Gregg Reynolds | 16 Jul 00:18 2007

Re: sedna ports?

Hi,

Thanks for the feedback.  Yes, OBSD has sys v shmget, mmap, etc., but
not the posix stuff (shm_open etc.)  I believe it does do address
space randomization.  If I have some time this summer I'll take a look
at the workaround you suggest, but I can't promise anything.  My free
time comes in bits and pieces, so it's hard to commit to something
that takes more than a few evenings or a weekend.  Well, ok, I have a
short attention span too, but don't tell anybody.

I'm not familiar with cmake, but I'm very familiar with gnu autotools,
so if you want to go in that direction I could certainly help.

Thanks,

gregg

On 7/15/07, ZNV <mejedi@...> wrote:
> Hello Gregg!
> It seems to me that OBSD has mmap sys call, doesn't it? Anyway the problem
> is more severe than simply switching to another IPC API. OBSD does address
> space layout randomisation. Sedna assumes she is able to map the shared
> memory segment at the same address in all transaction processes. Even worse,
> she assumes this address won't change from run to run. We need to redo many
> code to fix this. The best thing for now is a quick and dirty workaround if
> one exists, since the true solution is complex. Beeng able to map at least
> 10 Mb at the constant address (determined during installation for instance)
> will be enough, any ideas how to achieve this?
> And we still have no config stuff, shame on us :( The currently established
> solution works fine for Linux and Windows, however things must change if we
> are going to support more OSes officially. (I am advocating for CMake.)
> Things are done but sometimes very slow.
> Thank you for your interest in Sedna, we do really need your feedback.
>
> WBR
> Mejedi, Sedna Team
>
> 2007/7/14, Gregg Reynolds <dev@... >:
> >
> > Hi,
> >
> > I'm new to sedna and would like to run it on OpenBSD and maybe OSX.
> > Is anybody doing any ports work?  The shared memory stuff would need
> > to be reworked to support SYSV IPC, and it wouldn't hurt to tweak some
> > of the build/config stuff to improve portability.  There may be other
> > things but I've only just started looking at the code.
> >
> > Thanks,
> >
> > gregg
> >
> >
> -------------------------------------------------------------------------
> > This SF.net email is sponsored by DB2 Express
> > Download DB2 Express C - the FREE version of DB2 express and take
> > control of your XML. No limits. Just data. Click to get it now.
> > http://sourceforge.net/powerbar/db2/
> > _______________________________________________
> > Sedna-discussion mailing list
> > Sedna-discussion@...
> >
> https://lists.sourceforge.net/lists/listinfo/sedna-discussion
> >
>
>


Gmane