Data retention periods
DESK stores and retains different types of monitored data from your environments. The monitoring data is stored on the DESK Server. The following table shows the general retention periods for service data (PurePath), Real User Monitoring (user actions and user sessions), synthetic monitors, Log Monitoring, and metric timeseries data.
Data retention by type
Data type | DESK SaaS | DESK Managed | Storage |
---|---|---|---|
Services: Distributed trace and code insights | 10 days |
Configurable, with maximum 365 days of retention time |
Proprietary; Shared with non-aggregated RUM data |
Services: Requests and request attributes | 35 days |
Configurable, with maximum 365 days of retention time |
Proprietary |
RUM: Non-aggregated user action data (waterfall analysis, JavaScript errors, and crashes) | 10 days |
Configurable, with maximum 35 days of retention time |
Shared with distributed trace and code service insights |
RUM: Aggregated user action data | 35 days |
Configurable, with maximum 35 days of retention time |
|
RUM: User sessions | 35 days |
35 days |
|
RUM: Session Replay | Configurable, with maximum 35 days of retention time |
Configurable, with maximum 35 days of retention time |
|
Synthetic | 35 days |
Configurable, with maximum 35 days of retention time |
|
Log Monitoring | Configurable from 5-90 days. Specific files can be included/excluded. | Configurable from 5-90 days. Specific files can be included/excluded. | File-based NFS storage; Storage requirements and costs |
Timeseries metrics (Key user actions and requests) | Unlimited | Unlimited | Cassandra |
Services: Distributed trace and code insights
Includes PurePath data.
Services: Requests and request attributes
10 second granularity of charts, non-key and key requests
RUM: Non-aggregated user action data
DESK stores the full details of every user action for 10 days. This enables you to analyze individual user actions and get all details including waterfall analysis, JavaScript errors, and mobile crashes for 10 days.
RUM: Aggregated user action data
Aggregated user action metric (used in tables like Top user actions, Top JavaScript errors, and Top mobile crashes) are available for 35 days. After 10 days, user actions data is optimized for aggregated views and some individual user actions become unavailable for individual analysis. However the sample set is large enough for statistical correct aggregations.
RUM: User sessions
Includes Session Replay data. All user session data is stored for 35 days. Note that waterfall analysis, JavaScript error, and crash data are stored with RUM non-aggregated user action data.
RUM: Session Replay
Minimum size of required Session Replay storage volume is entirely load-dependent. A maximum size isn't required. In SaaS deployments, a dedicated disk is used for Session Replay data.
In Managed deployments, the Session Replay data storage directory is a dedicated file store that's used exclusively for Session Replay data.
Log Monitoring
Log Monitoring enables you to store all logs centrally within external storage. This makes log data available independent of log files themselves.
For DESK SaaS customers, log files are stored in Amazon Elastic File System in the zone where your DESK environment resides. You don’t have to worry about storage performance, availability, or free space. Disk storage costs are included in your Log Monitoring subscription.
To store log files centrally on your DESK Managed cluster, you must provide a common Network File System (NFS) mount point (path) that is identical and available from all cluster nodes. With this approach, it's your responsibility to ensure appropriate levels of performance, availability, and free space on the mounted NFS volume.
For full details, see Log Monitoring.
Timeseries metrics
0-14 days
: 1-minute interval granularity available for dashboarding and API access.14-28 Days
: 5-minute interval granularity available for dashboarding and API access.28-400 days
: 1-hour interval granularity available for dashboarding and API access.400+ days
: 1-day interval granularity available for dashboarding and API access.
Note: To provide accurate calculations for timeseries metrics, DESK uses the P2 algorithm to calculate the quantiles dynamically. This algorithm is known to yield good results and it works well with values in the long tails of value distributions. However, the aggregation algorithm is neither associative (i.e., (a + b ) + c == a + ( b + c )
) nor commutative (i.e., a + b + c == c + b + a
). This leads to slightly different response time estimates each time the algorithm runs. So, small differences in the response time values you may see in your environment (typically < 1%) are to be expected.