Forum Replies Created
-
AuthorPosts
-
peterlaursen
ParticipantI checked this in 7.12 RC2, and headers are written to log file. Not sure if such issue was there in 6.54. If there was we cannot do anything about it!
peterlaursen
ParticipantQuote:How do I do this?Data sync will let you sync from your remote to your local host. You will of course need to install a local server then! After sync you can backup from your local. To restore import to local and sync from local to remote. That is of course an 'overhead'. But it will work! And it runs unattended so maybe not a big problem!
Also maybe HTTP-tunnel would work around issues with configuration (that I still do not understand) so that you could back up from server directly. But even if it does not SJA from commandline will work also with direct connection.
Please start reading about HTTP-tunnel and Data Sync in the program help file!
peterlaursen
ParticipantNow MySQL has confirmed this as a server bug. This is not necessarily the final say about this, because now it will be forwarded to other people in the MySQL organisation.
Problem is that it also affects 4.1.22 and this one is out of maintenaince, so if and when they fix this it will probably never be fixed in 4.1 tree. So I think we will have to consider generating GRANT statement and not INSERT+FLUSH for column-level privileges.
However I do not consider column-level privileges very important (they are rarely used), so we will probably afford the luxury to wait for a few weeks more before making final decision about it!
peterlaursen
Participant'mod_security' is an Apache extension. Same would not happen with SSH tunnelling.
But I wonder if you have tried the 'urlencode' HTTP content-type (in HTTP advanced option in SQLyog connection settings). This will often solve issues wiht Apache 'mod_security'. You do not tell SQLyog/SJA version you use. It has to be a recent version to support this option.
peterlaursen
ParticipantOK .. same on 4.1.22
This illustrates the difference:
GRANT SELECT (ab_id) on test.abstracts to 'randomuser'@'%';
show grants for 'randomuser'@'%';
select * from mysql.columns_priv;
/*
Host Db User Table_name Column_name Timestamp Column_priv
% test randomuser abstracts ab_id 2008-10-22 19:31:56 Select
*/
— after dropping user
insert into mysql.columns_priv(host, db, user, table_name, column_name, column_priv) values('%', 'test', 'randomuser', 'abstracts', 'ab_id', 'Select');
flush privileges;
select * from mysql.columns_priv;
/*
Host Db User Table_name Column_name Timestamp Column_priv
% test randomuser abstracts ab_id 2008-10-22 19:33:41 Select
*/
show grants for 'randomuser'@'%';
/*
Grants for randomuser@%
GRANT USAGE ON *.* TO 'randomuser'@'%'
*/
I am almost sure that this is changed behaviour! And I am not able to understand why at least a server restart does not help this (I could understand if FLUSH failed)!
Let us see what they say here http://bugs.mysql.com/bug.php?id=40261
peterlaursen
ParticipantI think it is a MySQL server bug in recent 5.0 versions! I reproduced with 5.0.67. Try yourself something like this:
Code:insert into mysql.columns_priv(host, db, user, table_name, column_name, column_priv) values('%', 'test', 'randomuser', 'abstracts', 'ab_id', 'Select');
insert into mysql.columns_priv(host, db, user, table_name, column_name, column_priv) values('%', 'test', 'randomuser', 'abstracts', 'author_id', 'Select');
flush privileges;select * from mysql.columns_priv where user = 'randomuser';
/*
Host Db User Table_name Column_name Timestamp Column_priv
—— —— ———- ———- ———– ——————- ———–
% test randomuser abstracts ab_id 2008-10-22 18:49:10 Select
% test randomuser abstracts author_id 2008-10-22 18:49:10 Select
*/show grants for 'randomuser'@'%';
/*Grants for randomuser@%
————————————–
GRANT USAGE ON *.* TO 'randomuser'@'%'
*/Also after a server restart it is the same!
I will research and probably report to MySQL later!
peterlaursen
ParticipantCode:“and need unique IDs based on time”1) Well your current PK is a DECIMAL. Why not a DATETIME or TIMESTAMP PK then? that would work, but converting existing data may be a problem!
2) But also having all servers > 5.03 should solve this (if it is a problem wiht FLOAT). But how to ensure that existing stored DECIMALs are upgraded to new storage format in the upgrade process, I don't know (but MySQL like have it documented somewhere). I think safest would be to export and import again after upgrade.
3) Another good solution to this is to have a two-column PK of
* an autoincrement column
* a constant column with different values for each host
like
CREATE TABLE `tablename1` (
`id` int(11) NOT NULL auto_increment,
`worker` char(10) NOT NULL default 'joe', — use different defaults for each host
`blahblah` int(11) default NULL, — all your existing columns go here
PRIMARY KEY (`id`,`worker`) )
To me it looks like 3) is the most simple solution. And a PK based on time is not 100% safe anyway!
peterlaursen
ParticipantYou cannot have 'a second PK'. There can only be one PK. And creating a multicolumn PK will make no difference if one of the columns is a floating point column.
For one of the problematic tables try this
* 'duplicate table' on source (with structure and data)
* in the duplicate drop the currrent DECIMAL PK
* add an autoincrement INTEGER column to table and make a PK on that column
* now sync the duplicated table. First sync is no problem. I will simply copy to target (there is nothing to compare with).
* on source edit/add/delete a number of rows (still in the duplicated table)
* sync that table again
If it works now the problem is the FLOAT problem!
peterlaursen
Participantplease also tell:
1) The locale/language setting of your system
2) Did you use any special characters for naming folders or files?
We think it fails to populate the 'charset' dropdown of the dialogue!
peterlaursen
ParticipantYou are telling MySQL 5.0.51. But are both source and target that version? Note that before MySQL 5.03 DECIMALs are implemented as floting point values! Please make a test with a single table where you add an autoincrement PK!
peterlaursen
Participantlet us see tomorrow morning what information the dump has!
peterlaursen
ParticipantPlease tell the OS! Is it Vista?
If you can attach a sample CSV file where this is reproducable and a screenshot of the settings in the import dialogue that would be nice too!
added: please also paste the structure of the table you import to!
peterlaursen
ParticipantAs told in your ticket SJA data sync (from commandline) only will work on this server. This is the only solution we can offer if they insist on this configuration!
peterlaursen
Participantconclusion: one connection only per server with SJA data sync from command line.
peterlaursen
ParticipantConfirmed that for backup & copy db we are creating new connections. Probably also MySQL Administrator does for backup and fails for same reason!
They seem to disagree if also SJA (even from commandline) utilises 2 connections to same server. We are checking details in code now!
-
AuthorPosts