Is your feature request related to a problem? Please describe.
I'm always frustrated when I'm thinking of cluster restart (during upgrade, for example). This makes policers on all machines start at ~same time from ~same containers/objects. Which means a relatively high chance for simultaneous checks of the same objects from different machines. It's technically correct, but it's suboptimal wrt real synchronization process in case of netmap change, we want different machines trying different objects and restoring things faster overall.
Describe the solution you'd like
- Generate random CID/OID on policer init.
- Start listing objects from this pair.
- Treat cycle as finished once this CID/OID is passed again (comparing addresses).
Describe alternatives you've considered
Keep things as is.
Additional context
Mainnet and node management.
Is your feature request related to a problem? Please describe.
I'm always frustrated when I'm thinking of cluster restart (during upgrade, for example). This makes policers on all machines start at ~same time from ~same containers/objects. Which means a relatively high chance for simultaneous checks of the same objects from different machines. It's technically correct, but it's suboptimal wrt real synchronization process in case of netmap change, we want different machines trying different objects and restoring things faster overall.
Describe the solution you'd like
Describe alternatives you've considered
Keep things as is.
Additional context
Mainnet and node management.