This panel shows detailed information on the last x checkpoints, where x is 8 for older versions of Progress and 32 for 11.7+. Also, as of 11.7, the number of checkpoint statistics in memory is configurable via the -numcheckpointstats startup parameter.
|ProTop RT Label||Alertable Metric*||Description|
|Cluster||N/A||BI file cluster number of the checkpoint|
|Start||N/A||Start date and time of the checkpoint|
|Finish||N/A||End time of the checkpoint|
|Length||N/A||Length of time of the checkpoint|
|zcpLenSec||This alertable field is not displayed and contains the length of the ckpt in seconds|
|Dirty||dirty||Number of dirty buffers associated with the checkpoint|
|ChkptQ||cptq||Number of buffers in the checkpoint queue|
|Scan||scan||Number of buffers scanned by APWs during the checkpoint|
|APW Q||apwq||Number of buffers in the APW queue|
|Flushed||flush||The number of buffers written to disk at the end of the checkpoint. See below|
|Duration||duration||Duration of the chekcpoint|
|Synch Time||syncTime||Time to write modified database files to disk at the end of a checkpoint|
|BI Wr||biWrt||Number of writes to the BI file|
|BIWrTm||biWrtTm||Amount of time spent writing to the BI file|
|DBWr||DBWrt||Writes to the database|
|DBWrTm||DBWrtTm||Amount of time spent writing to the database|
*Requires data collector "Checkpoints" in pt3agent.cfg. With the commercial version of ProTop, you can configure alerts using the metric names in this column. For a full list of metrics, see the Alertable Metrics section.
Buffers Flushed Insight
At the end of a checkpoint, if there are still buffers remaining in the checkpoint queue not written to disk, the DB halts all forward processing until the remaining "flushed" buffers are written to disk. This value should always be 0.
There are three basic ways to respond to this problem:
- Improve your disk write performance. See the syncIO metric on the dashboard page
- Add more APWs
- Increase the BI cluster size
The first is the hardest to do, sincce infrastructure changes can be challenging and are often out of the control of the OpenEdge DBA.
The second is rarely the answer except on databases that do a huge number of writes. If you think you need more APWs, I'd challenge you to contact us first.
Increasing BI cluster size is the usual answer, especially if it's less than 16 MB. Note that the maximum BI cluster size is 256 MB.