The top SAP HANA alerts – revisited
Last year I wrote a blog post about the most common SAP HANA alerts. Many revision updates later I wanted to know whether the picture has changed. So I checked this time 42 production environment HANA instances and evaluated their alert data.
Of course, the meta-alerts 22, 23 and 24 (simply stating “some alert has occured”) are still at the top positions, being reported from virtually every system. Let’s ignore them and have a look at the real alerts. First of all, which SAP HANA alerts do occur on almost every system? This is shown by table 1:
| ALERT ID | ALERT TEXT | PREVALENCE |
| 0 | Internal statistics server problem | 88% |
| 58 | Plan cache size | 83% |
| 3 | Inactive services | 81% |
| 20 | Table growth of non-partitioned CS tables | 71% |
This is almost exactly the same picture as last year. While the SPS changed from mostly SPS 8+9 to now mostly SPS 10+11, the newly introduced alerts didn’t make it into the top positions.
The most significant change is the new default statistics server data retention. Starting in SPS 11, the statistics server only keeps 42 days worth of data (e.g. alert information). So for most instances only recent events were collected. Now let’s have a look at absolute number of occurences, providing a totally different picture, as shown in table 2:
| ALERT ID | ALERT TEXT | % of all events |
| 79 | Configuration consistency of systems in HSR setup | 42% |
| 43 | Memory usage of services | 12% |
| 21 | Internal event | 12% |
| 17 | Record count of non-partitioned CS tables | 8% |
While alerts 43 and 21 were already in that list on these positions last year, the old topscorer (alert 65) was replaced with a new topscorer (alert 79). Keeping the configuration of HSR systems identical on both sides is a real challenge, especially if irrelevant inconsistencies are being reported.
As a conclusion, alerts which happen on almost all systems (see table 1) have already some duplicate suppression mechanism implemented. And alerts which occur very often do not happen on many systems (see table 2), with the potential exception of alert 79 which happens on almost all systems with HANA system replication.