My fabric is 35 Switches, DCX, B48000, B5100, B5480 (Bladecenter module configured as switch) and so on
I know that the older switches limit the DB size to 1MB max but I am far from this limit, my actual cfgsize on the oldest switch model is below, so I'm 166 Kbytes in actual size, so far from the 1MB limit.
Does anyone alredy experienced this ?
FW versions running on the fabric are:
When I do a cfgsave, the following errors appears on the error log
2014/06/13-06:46:31, [RCS-1007], 1605,, ERROR, CTI_INT, Zone DB size and propagation overhead exceeds domain 191's maximum supported Zone DB size 127726. Retry after reducing Zone DB size.2014/06/13-06:46:31, [RCS-1007], 1606,, ERROR, CTI_INT1, Zone DB size and propagation overhead exceeds domain 193's maximum supported Zone DB size 127726. Retry after reducing Zone DB size.2014/06/13-06:46:31, [RCS-1007], 1607,, ERROR, CTI_INT, Zone DB size and propagation overhead exceeds domain 195's maximum supported Zone DB size 127726. Retry after reducing Zone DB size.2014/06/13-06:46:31, [RCS-1007], 1608,, ERROR, CTI_INT, Zone DB size and propagation overhead exceeds domain 197's maximum supported Zone DB size 127726. Retry after reducing Zone DB size.2014/06/13-06:46:31, [RCS-1007], 1609,, ERROR, CTI_INT, Zone DB size and propagation overhead exceeds domain 201's maximum supported Zone DB size 127726. Retry after reducing Zone DB size.2014/06/13-06:46:31, [RCS-1008], 1610,, ERROR, CTI_INT, Domain 191 Lowest Max Zone DB size
cfgSize ran on the Domain 191:Zone DB max size - 1045274 bytes
Available Zone DB size - 877904 bytes
committed - 166358
transaction - 0
When the zoning is distributed after cfgsave, the flash memory needs to allocate enough space for the cfg itself and also some kind of messing overhead, that I assume it becomes bigger as the number of switchs in the fabric grows.
In your case, it seems that due to the size of the fabric, the biggest cfg supported is 127726 Bytes (the rest of the space up to 1045274 Bytes has to be used for messaging). Since the cfg to distribute is 166358 Bytes in size, the fabric prevents you from distributing the zoning...
Thanks for your time!
Is there any way to "measure" this overhead ?
Looking to the supportshow output, i noticed a high usage of the root file system on the switches that generated the error. They are about 70% on the / filesystem.
When looking at your comments I would rather stick my head in a hornets nest than touch this fabric. 35 is not only scratching the certified limits of what Brocade has tested/supports but the sheer discrepancy of FOS code levels scares the heck out of me.
To get back to you question.
How many "Defined" configurations do you have in your setup? I know that when you use BNA for zoning actions it really easy to create a copy of an existing config, modify that one and then commit it to the fabric. Basically what happens is that a totally new config with all zones and aliases is added into the zone-database which could trip over the maximum size of the database of one of the switches.
As for your design I would seriously consider reconfiguring the baldecenter switches as an Access gateway and leave them out of the fabric. It becomes much more stable and far less risky.
Based on Brocade documentation, 35 is Far from the supported and tested limit in Broca. de Native Mode, wich is 56 from FOS 5 to FOS 7.
Also based on a Brocade analisys, they recommended to us to keep the bladecenter switches as a domain Member. That is to improve the ISL usage by better routing policy (in AG mode, the path is fixed in a blade-to-port basis and not balanced across the ISL's)
I totally agree with you, and I was scaried when started to work on this. We are planning a firmware upgrade initiative, but it is on customer's final decision.
Back to the main issue, I have only one defined configuration. and this occurs only when i try to add a simple zone / alias to a configuration through cli. On BNA i dont receive any error.
I don't know how to look deeply to it but i suspect on a problem o internal filesystem usage on the switchs reporting this error. Looking at supportsave, I can say that all switches that reported the error are 70% > on root file system