Maintenance Date and Time: August 7, 2017 From 2:30 p.m. to 4:30 p.m. (JST)
The switchover will be implemented while the system is online, but as a result of this the following circumstances will occur one or two times to the Mackerel main system during the time frame listed above.
- Unresponsiveness for less than 10 minutes.
- After the above, a period of about 20 minutes of instability.
The following will occur during the period of unresponsiveness.
- API access stoppage.
- Across-the-board failure of agent or mkr requests.
- Web screen access will also fail.
- Connectivity Monitoring alert stoppage.
- Subsystem stoppage (external monitoring, AWS integration, Azure integration).
The following will occur during the period of instability.
- Reduced response performance .
- Higher probability for access failures.
Effects of access failures.
In regard to the metrics posted by mackerel-agent or fluentd, these are buffered within the process itself and are later retransmitted, resulting in no data loss occurring. It is possible that there may be some loss related to metrics posted by independent systems such as cron.
In some cases, temporary issues have been observed in the graph data immediately after the system becomes responsive once more. However, the graph will display normally once the mackerel-agent buffer has finished retransmitting data.
During the time frame listed above, alerts will not send false alarms, but there is a possibility that the sending of alerts will not operate normally.
Regarding metric monitoring alerts, Mackerel’s system makes decisions on alerts when the metrics are received, so no false alarms will occur during the time frame listed above.
However, because mackerel-agent buffer retransmits in the order data was posted, it is possible alerts will not be sent in the case of certain settings, such when monitoring average values. In addition, in the case of check monitoring, alerts will not be sent when the API request fails.
About the announcement on that day
We will make an announcement before and after maintenance via the newsletter and the status page (http://status.mackerel.io).
Updates of the situation will be posted on the status page when necessary. In the unlikely event of the time frame for maintenance being extended, we will inform you via the status page.
In addition, after maintenance has been completed, we will post details on our blog, including the actual downtime.
Purpose of the Maintenance
There will not necessarily be new functionality added after a release, and this maintenance is in anticipation of scalability improvements, performance upgrades to the time series database, and functionality enhancement. In particular, there are plans to make changes to the time series database to allow for long-term archival of data with one-minute granularity.
For any inquiries regarding this matter, please contact us via email at firstname.lastname@example.org.
We thank you for your support of Mackerel and look forward to continuing to work with you in the future.