Upgrading is generally always safe (between many minor and one major version) and dashboards and graphs will look the same. There can be minor breaking changes in some edge cases which are usually outlined in the [Release Notes](https://community.grafana.com/c/releases) and [Changelog](https://github.com/grafana/grafana/blob/master/CHANGELOG.md)
Before upgrading it can be a good idea to backup your Grafana database. This will ensure that you can always rollback to your previous version. During startup, Grafana will automatically migrate the database schema (if there are changes or new tables). Sometimes this can cause issues if you later want to downgrade.
One of the database migrations included in this release will update all annotation timestamps from second to millisecond precision. If you have a large amount of annotations the database migration may take a long time to complete which may cause problems if you use systemd to run Grafana.
We've got one report where using systemd, PostgreSQL and a large amount of annotations (table size 1645mb) took 8-20 minutes for the database migration to complete. However, the grafana-server process was killed after 90 seconds by systemd. Any database migration queries in progress when systemd kills the grafana-server process continues to execute in database until finished.
If you're using systemd and have a large amount of annotations consider temporary adjusting the systemd `TimeoutStartSec` setting to something high like `30m` before upgrading.
If your using Grafana's builtin, LDAP (without Auth Proxy) or OAuth authentication all users will be required to login upon the next visit after the upgrade.
If you have `cookie_secure` set to `true` in the `session` section you probably want to change the `cookie_secure` to `true` in the `security` section as well. Ending up with a configuration like this:
```ini
[session]
cookie_secure = true
[security]
cookie_secure = true
```
The `login_remember_days`, `cookie_username` and `cookie_remember_name` settings in the `security` section are no longer being used so they're safe to remove.
If you have `login_remember_days` configured to 0 (zero) you should change your configuration to this to accomplish similar behavior, i.e. a logged in user will maximum be logged in for 1 day until being forced to login again:
The semantics of `max concurrent shard requests` changed in Elasticsearch v7.0, see [release notes](https://www.elastic.co/guide/en/elasticsearch/reference/7.0/breaking-changes-7.0.html#semantics-changed-max-concurrent-shared-requests) for reference.
One of the database migrations included in this release will merge multiple rows used to represent an annotation range into a single row. If you have a large number of region annotations the database migration may take a long time to complete. See [Upgrading to v5.2](#upgrading-to-v5-2) for tips on how to manage this process.
Pre Grafana 6.5.0, the CloudWatch datasource used the GetMetricStatistics API for all queries that did not have an ´id´ and did not have an ´expression´ defined in the query editor. The GetMetricStatistics API has a limit of 400 transactions per second (TPS). In this release, all queries use the GetMetricData API which has a limit of 50 TPS and 100 metrics per transaction. We expect this transition to be smooth for most of our users, but in case you do face throttling issues we suggest you increase the TPS quota. To do that, please visit the [AWS Service Quotas console](https://console.aws.amazon.com/servicequotas/home?r#!/services/monitoring/quotas/L-5E141212). For more details around CloudWatch API limits, [see CloudWatch docs](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/cloudwatch_limits.html).
Each request to the GetMetricData API can include 100 queries. This means that each panel in Grafana will only issue one GetMetricData request, regardless of the number of query rows that are present in the panel. Consequently as it is no longer possible to set `HighRes` on a per query level anymore, this switch is now removed from the query editor. High resolution can still be achieved by choosing a smaller minimum period in the query editor.
The handling of multi template variables in dimension values has been changed in Grafana 6.5. When a multi template variable is being used, Grafana will generate a search expression. In the GetMetricData API, expressions are limited to 1024 characters, so it might be the case that this limit is reached when a multi template variable that has a lot of values is being used. If this is the case, we suggest you start using `*` wildcard as dimension value instead of a multi template variable.
