diff options
| author | Brad Fitzpatrick <bradfitz@tailscale.com> | 2020-12-23 13:03:16 -0800 |
|---|---|---|
| committer | Brad Fitzpatrick <bradfitz@tailscale.com> | 2021-01-08 09:19:06 -0800 |
| commit | bc77c66290b9eee2c9050c6a96a0987a7f2904be (patch) | |
| tree | 8ded7495f60bb5114da5cf41092515d41a69e32d /internal | |
| parent | e692e3866bf2cbe240b64f96e03d590539e67aef (diff) | |
| download | tailscale-bradfitz/lite_endpoint_update.tar.xz tailscale-bradfitz/lite_endpoint_update.zip | |
control/controlclient: use lite map request handler to avoid aborting streamsbradfitz/lite_endpoint_update
Previously, any change to endpoints or hostinfo (or hostinfo's
netinfo) would result in the long-running map request HTTP stream
being torn down and restarted, losing all compression context along
with it.
This change makes us instead send a lite map request (OmitPeers: true,
Stream: false) that doesn't subscribe to anything, and then the
coordination server knows to not close other streams for that node
when it recives a lite request.
Fixes tailscale/corp#797
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Diffstat (limited to 'internal')
0 files changed, 0 insertions, 0 deletions