MySQL Downtime Issue

Question

Hi,  
  
I was experienced downtime with AWS RDS t2.medium instance for MySQL about 02:37PM to 02:41PM (UTC+8).  There was another downtime happened 2 days ago with exact same issue.  
  
I cannot find useful logs about the issue.  From the CloudWatch metrics in period of 1 Minute, I found during the downtime period, there are missing data for metrics like Freeable Memory, Write IOPS, Read IOPS, Queue Depth, Write Throughput, Read Throughput, Swap usage, etc.  
  
The most obvious changes on the metrics was Swap Usage, from 02:37PM with 132.867MB jump to 02:41PM with 224.016MB.  While 02:38PM to 02:40PM data are missing.  
  
Recently I change table from MyISAM engine to InnoDB engine 3 days ago due to table level lock issue.  
The CloudWatch showing  
Before changes:  
Freeable Memory around 1GB  
  
After changes:  
Freeable Memory around 100MB  
  
Is this the sign that I should go for higher instance, or possible other solution like parameter group adjustment, or OPTIMIZE TABLE after change of Database engine can solve them?  
  
Class  
db.t2.medium  
  
Region & AZ  
ap-southeast-1a

Accepted Answer

It sounds like you are short on memory. You can resize the innodb-buffer-pool-size to something lower than the default.  
  
When you switched to innodb, you started using the buffer pool more than before, the swapping is evidence that you ran out of memory.   
  
-Phil

Answer

Thanks Phil!  
  
These 2 days I focus on trying lower down the default innodb_buffer_pool_size to 2.5GB and 2GB.  The server seems more stable than before.  Default was 3GB, but when the issue happening I already lower down to 2.75GB.  During my research online, I always saw that should not make changes on the RDS default parameter group.  It seems in my case it need adjustment to lower down the memory usage to make server stable.    
  
Thanks again for pointing out and confirming the direction!

MySQL Downtime Issue

관련 콘텐츠