-
Notifications
You must be signed in to change notification settings - Fork 11
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove callmemaybe bits from compute #172
Conversation
@@ -71,25 +71,6 @@ zenith_connect() | |||
errdetail_internal("%s", msg))); | |||
} | |||
|
|||
/* Ask the Page Server to connect to us, and stream WAL from us. */ | |||
if (callmemaybe_connstring && callmemaybe_connstring[0] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
BTW it's possible to repurpose this code to help with old ps-1 version that requires callmemaybe.
Compute can send callmemaybe to the pageserver with the following content:
callmemaybe tenant timeline any_sk_address
It will trick old pageserver into connecting to new safekeeper and can work until we implement a more complex migration.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, that's what I've thought was possible and tried to advertise that a few days ago, but so far the consensus was to move on and purge/bump ps-1 after we notify external users about this event.
This way, we don't have to add anything at all in the code 🙂
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@petuhovskiy AFAIK In current setup compute does not send callmemaybe connstr to pageserver. It is the leftover of old setup that allowed to run without safekeepers, so should be ok to remove that entirely.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is the leftover of old setup that allowed to run without safekeepers, so should be ok to remove that entirely.
Yeah, I'm talking about hacking it to work in a new way.
but so far the consensus was to move on and purge/bump ps-1 after we notify external users about this event.
ps-1 on staging can be deleted, but on production we have to do migration somehow. Hacking callmemaybe in compute can be an alternative.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That would be an option, but direct compute->pageserver also has neondatabase/neon#1068 pit (apart from risk of losing data without safekeepers). So yes, let's do migration.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We can don't do direct compute->pageserver START REPLICATION, and just ping pageserver with safekeeper address, so old pageserver will know one safekeeper where it needs to connect.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ah, indeed
@@ -71,25 +71,6 @@ zenith_connect() | |||
errdetail_internal("%s", msg))); | |||
} | |||
|
|||
/* Ask the Page Server to connect to us, and stream WAL from us. */ | |||
if (callmemaybe_connstring && callmemaybe_connstring[0] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That would be an option, but direct compute->pageserver also has neondatabase/neon#1068 pit (apart from risk of losing data without safekeepers). So yes, let's do migration.
Looks like the migration is done differently, so merging this. |
Follow-up of neondatabase/neon#1619