[PDB-3684] PDB performance issues with large structured facts Created: 2017/09/20 Updated: 2018/02/13 Resolved: 2018/01/04
|Affects Version/s:||PDB 4.4.1|
|Fix Version/s:||PDB 5.2.0|
|Remaining Estimate:||Not Specified|
|Time Spent:||Not Specified|
|Original Estimate:||Not Specified|
|Sprint:||Hopper/Triage, Data Platform 2017-10-18|
|Method Found:||Customer Feedback|
|CS Frequency:||4 - 50-90% of Customers|
|CS Severity:||4 - Major|
|CS Business Value:||5 - $$$$$$|
|CS Impact:|| This is a very common and troubleshooting it has consumed a huge amount of support time. The workaround is to stop sending large structured facts which in some cases impacts their use of puppet.
In one example a custom structured fact collected rpm data which they were using in their code somehow so turning it off wasn't really a workaround but a reduction in functionality.
This can be very common on Solaris where there is a mount per user. A 20-30KB structured fact can be easily reached on solaris and cause this issue.
|Release Notes:||Not Needed|
As outlined in PDB-2631, PuppetDB encounters severe performance degradation in the presence of large structured facts. In support, we have seen this impact with mountpoint or partitions facts of only 20-30 KiB, which is not terribly uncommon. As noted in that ticket, the expectation is that PDB will be able to handle anything Facter might throw at it.
This issue was resolved in
|Comment by Nick Walker [ 2017/10/18 ]|
A configurable facts blacklist in PDB would mitigate but not solve this issue.
|Comment by Wyatt Alt [ 2017/12/06 ]|
We believe these issues will be resolved with https://github.com/puppetlabs/puppetdb/pull/2410
|Comment by Wyatt Alt [ 2017/12/13 ]|
The above has been merged.
|Comment by Nick Walker [ 2018/01/04 ]|
|Comment by Russell Mull [ 2018/01/04 ]|