Yes, but presumably then 255 routing tables with 1.5M routes each is also a relatively big deal. Perhaps some additional limit could be set to avoid such issues and allow a larger number of relatively simple rdomains with a handful of routes. Either way I take the 255 limit as the true and tested safe default.
Thank you both for your insight. Jun 9, 2026, 14:12 by [email protected]: > On Tue, Jun 09, 2026 at 01:05:04PM +0200, ptr wrote: > >> OK understand. Is a hypothetical 1023 still considered too large for this? >> > > 1023 routing tables with 5 route each, no big deal. > 1023 routing tables with 1.5 Mio routes each on the other hand is quite a > big deal. > > So there is no answer to this question. > >> Jun 9, 2026, 13:08 by [email protected]: >> >> > On 2026-06-09, ptr <[email protected]> wrote: >> > >> >> OK, thanks for the clarification. >> >> >> >> In my case rdomain IDs are sequentially generated and allocated, so this >> >> would be less of an issue with memory pressure. I believe the 255 limit >> >> could be on the low end if a larger fleet of nodes is factored in and say >> >> each tenant's identical rdomain IDs must potentially exist on every node. >> >> >> > >> > besides memory use, some operations use sequential scans of the various >> > rtables present on the system (i.e. at least one per rdomain), and won't >> > scale well with a larger number. >> > >> > >> > > -- > :wq Claudio >
