Authentication issues #13482
Labels
No labels
announcement
anubis
authentication
aws
backlog
blocked
bodhi
ci
cloud
communishift
copr
database
day-to-day
dc-move
deprecated
dev
discourse
dns
downloads
easyfix
epel
firmitas
forgejo_migration
Gain
High
Gain
Low
Gain
Medium
gitlab
greenwave
hardware
help wanted
high-trouble
koji
koschei
lists
low-trouble
medium-trouble
mirrorlists
monitoring
Needs investigation
odcs
OpenShift
ops
outage
packager_workflow_blocker
pagure
permissions
Priority
Needs Review
Priority
Next Meeting
Priority
🔥 URGENT 🔥
Priority
Waiting on Assignee
Priority
Waiting on External
Priority
Waiting on Reporter
rabbitmq
release-monitoring
releng
request-for-resources
s390x
security
SMTP
sprint-0
sprint-1
src.fp.o
staging
unfreeze
waiverdb
websites-general
wiki
Backlog Status
Needs Review
Backlog Status
Ready
chore
documentation
points
01
points
02
points
03
points
05
points
08
points
13
Priority
High
Priority
Low
Priority
Medium
Sprint Status
Blocked
Sprint Status
Done
Sprint Status
In Progress
Sprint Status
Review
Sprint Status
To Do
Technical Debt
Work Item
Bug
Work Item
Epic
Work Item
Spike
Work Item
Task
Work Item
User Story
No milestone
No project
No assignees
9 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
infra/tickets#13482
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Description of request
Our ipa cluster is fully reinstalled now, but there are still issues we need to sort out:
user facing/high priority:
internal/lower pri:
FYI: Retry your login
The FAS lockout threshold has been temporarily increased to allow people to retry their login until it works. The current threshold is 100 failed attempts in 10 minutes.
I have also readded all the servers to internal dns and redeployed noggin, so accounts may work more reliably now.
We are still looking for the root cause.
There is also IPA web UI authentication issue, do we want separate ticket for that or just add it to this one?
Oh, I was able to login now, not sure if I fixed that by #13486 or it is just a coincidence.
I have been able to login fine for a while, so I hope that issue is solved/gone.
I think everything is back to normal now... if anyone is seeing any problems, please let us know.
So, I can login here, but accounts.fedoraproject.org still says "Unauthorized: bad credentials." I have 2FA enabled, if that makes a difference.
😢
@abompard any ideas? can we get more verbose logs out of noggin?
Hi, I'm getting lately 400 Bad request every time I try to log in the first time. As I've observed, the authentication is successful but something in between the flow when tries to redirect back to forge goes wrong, getting the 400 response. Next try always is a 401 and the third time I click sign in, just acts as it nothing happens and let me in.
I also have 2FA
Ive tried with both Firefox and Chrome (with and without extensions)
I'm seeing the same behavior as flrnd using Epiphany Tech Preview with WebKitGTK 2.53.4. Most recently, just now:
This has been ongoing for several months now.
Hi @flrnd and @catanzaro The 400 issue is tracked in #13280
This is about accounts.fedoraproject.org login. Please lets keep issues in one place?
@kevin sorry! missed that 🙏
No problem at all... sorry if I came accross grumpy there... Hopefully we can fix this and that bug too.
I observed two things that are not yet documented, so I add them for the record, although I assume both have limited value:
while Firefox experienced the issue always, I could use accounts.fp.org in Falkon without issues. I tried in Firefox again while I was logged in with Falkon, but the issue remained in Firefox. It also remained after I logged out in Falkon.
after I deleted the cached files of firefox ("temporary cached files and pages" -> I did NOT delete "cookies and site data"), the issue was gone. We'll see if it comes back.
I use a default Firefox from our repos, up to date as of today, 153.0 (64-bit). Security settings are strict (but with "fix major site issues" enabled as recommended) and I have uBlock enabled (I do NOT use noScript at the moment:). Accounts used to work fine in the past with these settings, but I didn't use it for several months I think.
Since I cleared my cache, I can no longer reproduce the issue, but If there is anything specific to test or to document when the issue comes back, let me know.
I think I fixed the auth issue with IPA web UI. It seems the password max failures was set to 0, and locked user immediately after one failure. I set it to six and it seems to work.
Strange. I set that to 100 when we were having problems last week... unless it was a different setting, or somehow got reset?
It was set to 0, before I changed it.
Not sure if that IPA web UI issue is fixed. It's set to 100 again, but my account was locked when I tried to login and have to unlock it to be able to log in.
Yeah, so I don't think this is actually the problem. When people have said they cannot login I have run 'ipa user-status' on them and also 'ipa user-unlock' and it doesn't seem to matter.
Are folks still seeing this?
I think we need to either get noggin to log more about why it's saying unauthorized and/or match that up to a ipa dirserv log about the access. Would be nice to close this one out..
This is still happening. I tried today to log into Noggin and It wasn't working and then I waited a bit to try again and it worked and then I tried to log back in after my session timed out and couldn't again.
I think I found the problem on IPA.
This is what I can see in httpd error_log:
There are a lot of errors like this
Found that the
ipa-otpdwas failing withAug 14 08:59:56 ipa03.rdu3.fedoraproject.org ipa-otpd[3649437]: bind.c:089: Input/output error: IO error received on bind socket: Can't contact LDAP serverand restarted it which fixed the authentication issue I saw on IPA web UI.I unfortunately didn't found the root cause of this, but this would explain the
Bad credentialserror as the OTP didn't work correctly.Interesting. Does everyone who was seeing this have a otp enrolled? I think they might?
Note that ipa-otpd is setup as a systemd 'socket', so it listens for requests and then fires up the service. That service is sometimes getting that error. Explains why it might work after a while as the service times out and has to get started again from the socket.
I guess we need to file a ipa bug here and ask for help... I couldn't find anything already existing from a search.
All 3 ipa servers have those errors, so I don't think it's anything specific to ipa03.
And... I was able to hit the 'cannot login to the ipa web ui' issue... and otpd was working, it just rejected me:
So, not convinced this is the issue... :(
accounts.centOS.orgdoes not accept passwords of the length thataccounts.fedoraProject.orgdoes. #13507Yes, I have OTP enabled (but not sure if I'm still affected as I didn't use accounts.fp since my last post)
In case of IPA web UI it usually helps me to just do
ipa user-unlock, but I'm not sure if accounts.fedoraproject.org is the same problem and I'm also not sure why the account get locked in first place.EDIT: One idea crossed my mind, isn't there some grace period after giving the wrong login credentials before it will let you try again.
I have OTP enrolled.
This could be exacerbating things, though there has definitely been some amount of false "bad credentials" happening. I use 1Password, so zero opportunity to typo anything. (Unless someone is trying to brute force my password and locking my account! Clankers?)
So, I went down this path too, but everytime I (or someone else) was having problems logging in and I checked with 'ipa user-status ' it said they were not locked, and 'ipa user-unlock' didn't seem to do anything. ;(
As far as the time limits, I set them to 10 seconds a while ago, and the number of failed logins is 100...
So, if it is this thing, it's acting in a way that doesn't seem as described.
Do we know for sure if this issue only affects users with 2FA enabled or if that's a red herring? If this is 2FA related, it would be ideal to have it sorted before the provenpackager 2FA flag date at the end of September.
(EDIT: If it helps, I also have 2FA enabled and hit this again today)
We are not sure, but it looks like a common pattern.
I can log into accounts, but I can no longer log into id.fedoraproject.org or id.centos.org since enabling 2FA OTPs. I now persistently get 401 errors.
Huh, thats different and the first report we have seen of that. By "log into id.fedoraproject.org or id.centos.org" you mean you try and log into some application, you don't have a kerberos ticket and it redirects to id to ask your credentials?
What application was it?
What error did it give? just incorrect password?
If you do have a valid kerberos ticket does it log you in?
@kevin wrote in #13482 (comment):
Yup.
Three applications so far: RHBZ. GitLab.com. Dist-Git.
I get "authentication failure" today, but yesterday I got "unauthorized".
Yes, but this is not applicable on my phone or my Mac.
So, this thing turned out to be sssd on ipsilon01 being in a bad state. I am guessing that it worked for people who it had cached info for, but not for anyone else. :( I think it was caused by the ipa02 re-sync, but not sure of that. In any case, restarting sssd on ipsilon01 seems to have fixed that.
(But will not I am pretty sure fix the sporadic problems logging into accounts / ipa web ui)
So I reinstalled ipa02 and ipa03, hopefully that will make the ipa cluster more stable. Let's see if that helps with this issue as well.
I've tweaked some things on ipsilon01 to avoid the bad state ( I hope ).
If folks see any issues now, can you please let us know? (and which exact issue?)
Reminder, the issues are:
I've seen no reports in the last 5 days. If anyone is still seeing something, please let us know.
I'll keep this open for a bit more to gather more issues if any.
Maybe it helped when I reinstalled the ipa02 and ipa03 again with you tweaking the ipsilon01.
Yeah, I think so...
I guess I am going to close this now.
Please report any new issues.