You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently mirrobrain (MB) is running on the storage node which implies:
any load (cpu, mem) to the MB load-balancer affects storage node
any I/O induced by MB (postgres database queries) affects storage node
any bandwidth consumed by MB (numerous small queries mostly: redirects to mirrors and MB-specific endpoints like btih, torrent, magnet, metalink)
should storage node (ie. master mirror) be unavailable for some reason, the whole download.kiwix.org is down while all mirrors are probably still up and ready
It would thus be better for performances and reliability to have the services node host the download.kiwix.org mirrorbrain.
The text was updated successfully, but these errors were encountered:
I think all this comes down to the question of whether it is possible to have mirrorbrain operate without access to "real" files. This has been discussed in the HA discussions, and still needs to be checked AFAIK. If the conclusion is positive, I agree it would be positive to move MB to another node AND have an active or passive backup instance.
Currently mirrobrain (MB) is running on the
storage
node which implies:storage
nodestorage
nodebtih
,torrent
,magnet
,metalink
)storage
node (ie. master mirror) be unavailable for some reason, the wholedownload.kiwix.org
is down while all mirrors are probably still up and readyIt would thus be better for performances and reliability to have the
services
node host thedownload.kiwix.org
mirrorbrain.The text was updated successfully, but these errors were encountered: