MySQL Bugs: #102030: Segmentation fault while Load with shell util.loaddump()

Bug #102030	Segmentation fault while Load with shell util.loaddump()
Submitted:	18 Dec 2020 20:46	Modified:	23 Dec 2020 20:32
Reporter:	luis garcia	Email Updates:
Status:	Not a Bug	Impact on me:	None
Category:	MySQL Server	Severity:	S3 (Non-critical)
Version:	MySQL Shell 8.0.22	OS:	CentOS ( 3.10.0-1160.6.1.el7.x86_64)
Assigned to:	MySQL Verification Team	CPU Architecture:	x86 (x86_64)

Description:
Hello

While I use this:

mysqlsh --mysql -u root -h localhost

\sql set global local_infile=1;
\sql SET GLOBAL log_bin_trust_function_creators = 1;

util.loadDump("/var/lib/mysql/nfs/backup", { threads: 8})

I have the folowing error:
8 thds loading - 6% (1.70 GB / 27.64 GB), 426.42 KB/s, 0 / 22138 tables doneSegmentation fault

And from /var/log/messages I have this:

Dec 18 21:32:47 vmcosdevdb01 kernel: mysqlsh[17453]: segfault at 7efff3f2937f ip 0000000000f2048c sp 00007efffbffdbb0 error 4 in mysqlsh[947000+11eb000]

Many thanks if you can resolve this issue.

How to repeat:
Not sure you can reproduce this without my databases

Hi,

Are you connecting directly to MySQL server or trough MySQL Router?

We had some issues with 8.0.22, can you try to load the same dump with 8.0.21 to see if this will solve the problem?

Thanks
Bogdan

Hello Bogdan

thanks for your update.
So I remove shell 8.0.22 from source and slave
and I install 8.0.21 from RPM file
And I redo

from source:
util.dumpInstance('/mnt/mysql_migration/backup',{ threads: 8})

and from replica

mysqlsh --mysql -u root -h localhost

\sql set global local_infile=1;
\sql SET GLOBAL log_bin_trust_function_creators = 1;

util.loadDump("/var/lib/mysql/nfs/backup", { threads: 8})

And after a while I have this error:

Util.loadDump: Cannot open file '/var/lib/mysql/nfs/backup/citobi6979@DataSource_6f04ca2a_1859_4f30_b985_c3d3cd0bff68.sql': Too many open files (RuntimeError)

Let me know
Thanks

Hello
and the question:
Are you connecting directly to MySQL server or trough MySQL Router?

It's direct conexion, no router in the middle

I have this parameter 
in  /etc/security/limits.conf

mysql hard nofile 65535
mysql soft nofile 65535

hello

Ok I increase ulimit parameter and then I redo

Now I have this error:

Util.loadDump: Create table/tablespace 'portal_portal_account_folder_item' failed, as disk is full (MySQL Error 3675)

The / is full
But database datadir is outside /
So, what is filling / partition ? Can you tell me?
Many thanks

Hello

Ok I see that we have some free space issue. Once this issue have been solved, now I have this error:

util.dumpInstance('/mnt/mysql_migration/backup',{ "showProgress" : "true" })
Util.dumpInstance: Unknown database 'citobi745363' (RuntimeError)

Let me know what I have to do.
Many thanks

the problem was related to previous dump with non existing data anymore.

Hi Luis,

I cannot reproduce this. The full disk, open files etc. are config issues that I see you figured out how to solve (note on systemD setup you need to setup open files limit in the systemd service file for mysqld too). 

Now we have some issues in .22 with "large packets" but mostly in router. Few bugs in .22 too but I don't see they are related directly to what you are experiencing but asked to check with .21 just to be sure. Now I restored a 34GB dump from nfs into MySQL Server doing the same thing you did and I had no issues. Restored also few smaller dumps (5GB and 1GB). Might be related to your data, can't confirm without data and I assume you can't share that.

Please clarify what happened with .21 testing (not sure I understand F.D. comment) to see how to proceed.

all best
Bogdan Kecman

The user was also looking for help on MySQL Community Slack.

The problem was that he was taking an inconsistent backup and at the same time removing databases.

Hi Frederic,

Thanks for the clarification. Yes, in that case, there's no bug really.

All best
Bogdan