Ubuntu 18.04.2 LTS

Column 2 is 8176 bytes which happens to be the block boundary size for TEXT data (16 bytes for block header). I have a feeling block allocation is looping in Dctnry::insertDctnry trying to find a suitable block and not finding it.

I have sql dumps with lots of inserts. Do you have any idea how to handle or avoid those unfortunate rows? I got a little bit stuck.

Hi arcturus, unfortunately I don't yet know of a good workaround beyond making sure the length of data for a TEXT column isn't divisible by 8176 bytes.

Alternatively you could store the text in an InnoDB table and join as required (this may be faster anyway). Or I think cpimport may not have the same bug (I have not yet checked).

Thank you. Cpimport (used from command line) successfully managed to insert this row and all others (from csv file). Although I had to change data types from blobs to texts.

I have encountered another problem with importing my data to another table. Now my csv file size is 130GB and I'm trying to cpimport it the same way like the previous file (7GB). Despite the setting in Columnstore.xml <NumBlocksPct>50</NumBlocksPct> and even lower (I restarted server too) my memory goes to max (~15GB). Next swapfile also goes to max (2GB) and then server restarts.

I got following errors:
Local Memory above Critical Memory threshold with a percentage of 100 ; Swap 100
Local-Memory usage at percentage of 100 , Alarm set: 7
Memory Usage for Process: nautilus-desktop : Memory Used 2796 : % Used 1
Memory Usage for Process: Xorg : Memory Used 5936 : % Used 1
Memory Usage for Process: gnome-shell : Memory Used 8816 : % Used 1
Memory Usage for Process: mysqld : Memory Used 175819 : % Used 11
Memory Usage for Process: cpimport.bin : Memory Used 1373310 : % Used 85
Swap above Critical Memory threshold with a percentage of 100
Swap usage at percentage of 100 , Alarm set: 10
Swap Space usage over Major threashold, perform OAM command restartSystem

Turning off swap memory didn't fix the problem.

Sorry that I mention my next problem here but I don't feel like it deserves new topic. Please correct me if I'm wrong.

arcturus actually if you could create a new Jira ticket for this it would be great. Along with details of your setup (number of UMs and PMs). Looks like we may not be freeing memory in cpimport for TEXT/BLOB correctly which is a different problem.

Verified on,
Build: 1.2.6-1
gitversionEngine: 15aea637

'insert4.sql' works fine on the above build.

Ex:
root@vagrant:~# mcsmysql test < insert4.sql
root@vagrant:~# mcsmysql test
MariaDB [test]> select count from text;
----------

count

----------

----------
1 row in set (0.062 sec)