[vygi]

Hello there,

 

we have XP Professional with 4GB and integrated graphics which uses 1GB. 3GB of RAM are available for the OS. Max. 40% (1.2G:cool: used and at least 1.8GB free, even when additional applications are running and SQLyog is holding and showing results of 20-30 executed (mostly vers light) queries. Then it needs 50-60MB of RAM. Never more.

 

Now it crashed again. Ca. 2,006,000 KB of RAM (almost 2G:cool: free AFTER restart or SQlyog. The SQLyog itself currently uses 38MB.

 

This issue came together with the SQLyog sessions.

It never crashes during query execusion or other operation. It always hapens in the background, or when I switch back to the SQLyog with Alt-Tab.

[vygi]

I wish we could upgrade… our sysadmin said “No” 🙁

 

At least two of my colleagues have same problem.

Fortunately, the session (all connections and all query tabs) is being always fully restored, so that’s not a real problem. Just kinda annoying…

 

Thanks for your support, Peter!

[vygi]

No; nothing special what could be related to SQLyog, and no events at all at the time the yog has crashed…

 

I will ask our sysadmin to upgrade my workstation to Windows 7 as XP is really too old…

 

Anyway, another screenshot attached; task manager is now visible.

 

[attachment=1978:SQLyog_crashed2.PNG]

 

As you can see there are no memory allocation changes within the last minutes.

Slightly less than 1000 MB of RAM was used (with SQLyog) yesterday all the time. I’ve locked up my workstation in the evening. Now this “Not enough memory” message and 1.01 GB used. It dropped to 923 MB after I clicked “OK” and now 951 MB after the SQLyog is restarted (session restored but no queries executed so far).

[vygi]

here we go:

[vygi]

Thank you;

 

my PC has 4 GB RAM in total.

3.319.748 KB available for the 32-bit OS,

right now only 1,04 GB (1/3) used and 2.210.736 KB free.

 

More info and screenshots comes as soon as the program crashes (this or next week).

 

Vygi

 

 

 

As a first step could you please attach a screenshot of the error dialog? This will help us to understand the context (if it is a Windows dialog or a SQLyog dialog returning this error for instance). 

 

Also please tell how much RAM you have in the machine and how much is typically used/free.

[vygi]

Yes, thanks; I thought it was working in previous verrsion, but it was not.

Well…. let's wait and drink some coffee in the Christmas time 😉

[vygi]

peterlaursen wrote on Mar 12 2009, 09:44 AM:

I found this:

http://bugs.mysql.com/bug.php?id=35824

.. but according to this it should have been fixed in 5.1.25.

Anyway I think you should start locating the statement (from HISTORY) causing this. Once the statement (or sequence of statements) has been identified it should be reproducable with all clients.

Thanks;

this error comes even after simple SHOW EVENTS.

I have contacted MySQL support and submitted this issue.

[vygi]

peterlaursen wrote on Mar 11 2009, 05:33 PM:

Not reproducable for me with server 5.1.32.

Note there error message “Cannot proceed because system TABLES used BY Event Scheduler were FOUND damaged AT SERVER start”. I remember this error from bugs.mysql.com. I am not sure if if mysql_upgrade program will fix this (hopefully it will). The error message is not ours. It is the server that sends it!

The problem is that the `mysql`.`event` is damaged or incorrect. Was mysql_upgrade executed after last upgrade? Maybe the table does not exist (will happen if server was upgraded from a 5.0 version without running mysql_upgrade.

So 99.9% not our issue. Upgrade the server the correct way or complain to MySQL!

Hello and thank you;

the server was upgraded acording to mysql recommendations: 5.0 RPMs completely removed, then 5.1 installed, and mysql_upgrade executed.

I will contact MySQL to ask what's wrong.

Thanks,

Vygi

[vygi]

peterlaursen wrote on Mar 11 2009, 01:46 PM:

OK .. so you use the hostname/name server alias 'bee'. The reason I asked was that there could be a DNS-'broadcast' problem with Samba or with the network. If that was the reason using the ip instead of 'bee' would probably be faster.

(I remember some years ago that when I had a network with 2 Windows machines and one SuSE 10 (running both Samba server and client), I had lots of networking problems with hostnames if the SuSE firewall was enabled. Disabling this firewall or using the ip instead of host name always worked fine.)

Do you know if it is Samba that broadcasts the hostname or do you have a centralized name server on the network?

You can reply to those if you know the answers. But we will need to try this ourselves. Our test team will be fully manned from tomorrow – it has not been for the last week.

We have a DNS server.

I have tried to write to \192.168.1.70profilesoutput.txt instead of \beeprofilesoutput.txt but with same result: it's not faster.

Thanks,

Vygi

[vygi]

peterlaursen wrote on Mar 10 2009, 03:11 PM:

ok .. we will need to set up a test environment for that. Can you tell how you specify the machine where this shared folder is? Using the ip or a hostname/alias?

This is my home directory (/home/vygi) on a Linux server “bee” which has Samba.

On the Windows machines, everybody's home directory is visible as \beeprofiles and I use my own as network drive b:

So when I export data to b:output.txt (which is same as \beeprofilesoutput.txt) it is being (sloooooooowly) saved on “bee” as /home/vygi/output.txt.

Hope it's clear enough?..

Thanks,

Vygi

[vygi]

peterlaursen wrote on Mar 6 2009, 11:45 AM:

You do not tell the version you are using currently!

Changelogs for 8.02 “CHUNK setting (for exports) now also has effect for non-HTTP-tunneled connections. Using this option is required if it takes more time to retrieve data from a table than server ‘net_write_timeout’ setting.”

Could you try 8.02 or 8.03 and define a CHUNK setting in 'preferences' and see if this makes any difference for speed of file store?

Sorry…. we have 8.02 here but “chunk size (in rows) during batch exports” does not make our exports to network drives any faster.

[vygi]

peterlaursen wrote on Mar 6 2009, 11:09 AM:

I think we had a similar report before. We were not able to reproduce. But we probably did not try with shares created with SAMBA.

We will discuss a work-plan for researching into this. There may be several variables – SAMBA version, configuration etc.

Thanks for the quick feedback;

we have tested several network drives from several Samba servers with same result: they're 50-70 times slower than local discs.

Clients: Windows XP Professional

Servers: Linux SuSE 10 and RedHat Enterprise 4 with more or less new SAMBA versions.

Regards,

Vygi

[vygi]

Ooops…. wrong forum! Sorry!

Moderator/admin: please “move it move it”!

[vygi]

peterlaursen wrote on Jun 6 2007, 12:32 PM:

One Query tab and one Query Builder tab will be opened in every new connection.

We did not consider making this configurable!  Is it so important?

No, of course it's not so important, just a bit annoying and feels like made by Microsoft 😉 I have to close it every time, because I never need it.

[vygi]

Thanks;

and I just want to say that I can perfectly reproduce it like described above with two MySQL 5.0 servers and SQLyog 6.0 RC2, on German WinXP Professional SP2.

[Author]

Posts