[cloudcontrol]

While we have received alerts when we replication is X seconds behind the master, the error state we encountered Sunday (see attached history export) did not trigger an alert. We find it hard to believe our error condition is not by default one that would generate and email notification.

The circumstances where that the replicas were reporting the following error: Got fatal error 1236 from master when reading data from binary log: 'Client requested master to start replication from impossible position'

We know that this was caused by a RDS master failover on AWS, abd as you can see from the moyog report, replication was completely broken. How can we make sure MONyog alerts us if this happens in the future?

Thanks!

[cloudcontrol]

'Mahesh' wrote:

You may refer this blog post:

http://www.webyog.com/blog/2011/07/20/what-are-hardware-requirements-for-monyog/

Thanks- that is one of the better requirements pages I have seen. Seems like we should be OK on resources, but I went ahead and created a RAID 0 array of 10 EBS volumes just to make sure. I have the monyog root symlinked to a directory on the striped array and still have not seen a performance increase.

Unfortunately while the tool is adequate for near real time information, it is completely unreliable when I try to look at historical data. Sometimes it works, and other times I have experiences like today: I have been trying for over two hours to view data from a simple Monitors/Advisors History/Trends page for the range 6:30 AM to 7:15 AM on a single server. Only stopping monyog and restarting monyog will allow me to bring the page up completely. Once up, I have twice waiting another 45 minutes for a details page which would give me a minute by minute breakdown of insert statements during that historical period. The RAID has not caused this issue- this problem existed before moving the monyog root folder as well.

Attached is a screenshot of the monyog systems' “htop” command and a cloudwatch screenshot. The cloudwatch screenshot is most interesting- it shows we have excess CPU that is not being utilized, and that disk iops are plateauing at 200/sec. Cloudwatch cannot show us disk iops on the RAID array, unfortunately. There is also 5 GB of RAM available on the system, which is also used for very lightly accessed mediawiki and joomla installation.

Question I have are:

1- is monyog using temp space for reads writes outside the monyog root (which is now running on the raid 0 array?

2- why won't monyog “grab” more system resources?

3- given the specs you linked to, or usage seems quite light- why can't I reliably load that simple historical page?

Thanks for your help!

[cloudcontrol]

'Mahesh' wrote:

which application parameters might decrease the disk i/o ops ?

— I think you here you can increase Data collection interval to 5 minutes

which is default?

Also why you want to have Data collection = 10 second on a Production server ?

If you increase to 5 minutes definitely you will see significant decrease on Disk I/o operations.

Thank you- 5 minutes is a bit too slow to react to problems on our social networking site. We are now using 1 minute as the interval and have seen performance improve, though still slow to “load up” the monitor/advisor page initially. Regarding load MonYOG puts on our database: I am not sure I understand your concern about a 10 second interval on a production server- is monyog performing selects or simply monitoring the process list? Given the hardware specs I posted, would you imagine a 10 second interval is seriously taxing the system?

'Mahesh' wrote:

Also another point is consistently viewing Monitors/Advisers page, Refreshing it frequently?

You can of course want to see stats about your server at any time,viewing less frequently or refreshing page frequently might decrease Disk I/O operations.

MONyog resolves each counter when you are viewing on Monitors/Adviser page.

We are working on optimizing resolving MONyog counters.

I see. No, we are not refreshing, because each page refresh takes a great deal of time to load all of the counters in monitors/advisors and we prefer to watch the counters update automatically.

Thanks for your help!

[Author]

Posts