Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


HostHatch Los Angeles storage data corruption (Was: HostHatch Los Angeles storage down)
New on LowEndTalk? Please Register and read our Community Rules.

HostHatch Los Angeles storage data corruption (Was: HostHatch Los Angeles storage down)

Daniel15Daniel15 Member
edited May 6 in Outages

My storage VPS went down at 7:04 PM PDT (as I write this it's 9:28 PM). It just shows a loading spinner if I try to access it in their control panel:

Opened a ticket and Mike says they're working on it.

I hope it's not involucrated like Chicago (https://lowendtalk.com/discussion/comment/3395311).

«134567

Comments

  • dahartigandahartigan Member

    @Daniel15 said: involucrated

    Let's hope not.

  • nyamenknyamenk Member

    I can confirm my Storage LAX is up

  • hotsnowhotsnow Member

    mine also offline :'(

  • Daniel15Daniel15 Member

    @nyamenk said:
    I can confirm my Storage LAX is up

    I wish they showed the node name in their control panel so we could see if we're on different nodes.

    Thanked by 2nyamenk bdl
  • cablepickcablepick Member

    Same for me. Just came here to see if it was affecting anyone else too.

    I was on the Chicago server that involucrated and I made this LA server my primary, with backups to Chicago and London.

    Should I post the rest of my server in case anyone wants to move away from them since apparently I'm bad luck?

    Thanked by 1Logano
  • hosthatchhosthatch Top Host, Host Rep

    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Thanked by 1Frameworks
  • cybertechcybertech Member

    mine is up too

  • hotsnowhotsnow Member

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    seems no any email notifications

  • kennsannkennsann Member

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Sadly mine's down too, I dont usually check my storage server since its kinda "cold" storage for the backup of files. I only do check it when I upload new stuff for archival.

    Hearing this in public first rather than the provider through email or other means is a bit concerning.

    Thanked by 1Logano
  • Daniel15Daniel15 Member
    edited May 5

    Came back online at 9:49 PM but then went down again at 9:55 PM (PDT)

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Because you don't often send email notifications while outages are happening, the control panel doesn't display announcements about outages, I wanted to see if it affected other people (since sometimes outages were isolated to just my VPS), and I was worried it'd be another big event like Chicago.

    It's not uncommon to have users create threads during outages, although it seems like it was more common on WHT compared to here :smile:

  • risharderisharde Member

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Why shouldn't there be a thread? This is the second time you seem to be down with issues. These threads are helpful for people like me and this would have been an issue for me if I was a customer. I will hold back from purchasing storage until things seem more safe on your side of things.

  • Daniel15Daniel15 Member

    Mine's back but needs a manual fsck

    but I'm about to go to sleep so I'll deal with it tomorrow morning. Hopefully it's not broken to the point that fsck won't fix it.

  • dahartigandahartigan Member

    @hosthatch said: Not sure why there needs to be a thread, but sure :)

    You're supposed to think that but not say it ;)

  • hosthatchhosthatch Top Host, Host Rep

    @risharde said:

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Why shouldn't there be a thread? This is the second time you seem to be down with issues. These threads are helpful for people like me and this would have been an issue for me if I was a customer. I will hold back from purchasing storage until things seem more safe on your side of things.

    Just to be clear, one node was down, which is about 5% of our storage users there, and not all storage in Los Angeles.

  • cybertechcybertech Member

    @Daniel15 said:
    Mine's back but needs a manual fsck

    but I'm about to go to sleep so I'll deal with it tomorrow morning. Hopefully it's not broken to the point that fsck won't fix it.

    this happend to mine too, about 1-2months ago

  • hosthatchhosthatch Top Host, Host Rep

    @cybertech said:
    this happend to mine too, about 1-2months ago

    It's a fairly common thing to see forced fsck's after a hard reboot on large disk machines.

  • plumbergplumberg Member

    How does one fix the same? I am noob on this so will gladly appreciate some help.

    Centos7 --> FDE setup. Have entered the LUKS password and it has spit out some weird errors. Not sure how to proceed. thanks

  • kevertjekevertje Member

    Dear diary,

    Fsck!

    Thanked by 1TimboJones
  • @plumberg said:
    How does one fix the same? I am noob on this so will gladly appreciate some help.

    Centos7 --> FDE setup. Have entered the LUKS password and it has spit out some weird errors. Not sure how to proceed. thanks

    What's the error?

    If it's a damaged LUKS header, you'd better kiss that data goodbye since you're not getting any of that back if you don't have a good copy of the header.

  • causecause Member

    verified from livecd and found many checksum errors. For those who could mount the corrupted storage, do not forget to verify the contents.

  • LiliLabsLiliLabs Member

    @hosthatch said:
    Just to be clear, one node was down, which is about 5% of our storage users there, and not all storage in Los Angeles.

    I love you guys, but those 5% of users should really have been notified. I think adding node name into the panel + sending outage emails would go a long way.

  • darkimmortaldarkimmortal Member
    edited May 5

    It's a fairly common thing to see forced fsck's after a hard reboot on large disk machines.

    It’s not unless there is some hardware issue, or some layer of the storage setup is defeating fsync etc. With a block device that holds data and fsyncs correctly, it is always just a matter of replaying journal on hard reboot. A modern filesystem will never be corrupted in this scenario

    We've already resolved this problem by design in the new nodes we have set up in the past two years. This was one of the older nodes.

    https://lowendtalk.com/discussion/comment/3397397#Comment_3397397

    Is it this issue again?

    Thanked by 2stevewatson301 adly
  • skorousskorous Member

    @LiliLabs said:

    @hosthatch said:

    I love you guys, but those 5% of users should really have been notified. I think adding node name into the panel + sending outage emails would go a long way.

    status.hosthatch.com doesn't even show having had an outage. Is that no longer a valid way to find out for myself whether things are having a problem?

  • Daniel15Daniel15 Member
    edited May 5

    I ran fsck.ext4 and there was a bunch of corruption which is a bit concerning. I've got around 4000 inodes in lost+found now, across both the root volume group and data volume group :/

    So far I've noticed that both the Syncthing and Netdata executables are corrupted, as are several of the cached apt package lists. apt update failed to run because several of the lists had a chunk of the systemd journal in them 🤔

    I'm working out if I can verify the files based on my Borgbackup backups (borg check?), otherwise I'll just restore everything from backup.

  • jugganutsjugganuts Member

    Sad too hear yet another data loss story from hosthatch... whats that 3-4 in the last 2 years... (that iv heard about)

  • Daniel15Daniel15 Member

    Most of the important data seems fine, which is good. The system booted fine so all the core files required to boot are there. I'm probably still going to restore everything (~5TB) from backup.

  • risharderisharde Member

    @hosthatch said:

    @risharde said:

    @hosthatch said:
    Not sure why there needs to be a thread, but sure :)

    We're working on it.

    Why shouldn't there be a thread? This is the second time you seem to be down with issues. These threads are helpful for people like me and this would have been an issue for me if I was a customer. I will hold back from purchasing storage until things seem more safe on your side of things.

    Just to be clear, one node was down, which is about 5% of our storage users there, and not all storage in Los Angeles.

    Since op seemed to have recovered from the issue, I don't want to get deeper but I think knowing there was an outage was beneficial for me.

  • darkimmortaldarkimmortal Member
    edited May 5

    @Daniel15 said:
    I ran fsck.ext4 and there was a bunch of corruption which is a bit concerning. I've got around 4000 inodes in lost+found now, across both the root volume group and data volume group :/

    So far I've noticed that both the Syncthing and Netdata executables are corrupted, as are several of the cached apt package lists. apt update failed to run because several of the lists had a chunk of the systemd journal in them 🤔

    I'm working out if I can verify the files based on my Borgbackup backups (borg check?), otherwise I'll just restore everything from backup.

    Looks like another case of corruption of data at rest... very weird failure mode. Maybe understandable if that systemd log data was recent, but "Jul 14"?

    For borg the best bet is probably to mount the latest backup and use another tool to compare. I don't think it has any option to force checksum

  • Daniel15Daniel15 Member

    @risharde said: Since op seemed to have recovered from the issue

    Working on it... there's some data corruption.

    @darkimmortal said: and use another tool to compare.

    Since the backup is encrypted, it'd have to download and decrypt all the data to compare it. At that point I may as well just restore the entire backup anyways. I realised that I actually only have ~1TB in the backup as most of the space used on the server is actually backups from other systems where I can just use borg check to check it.

    Thanked by 1risharde
  • darkimmortaldarkimmortal Member
    edited May 5

    @Daniel15 said:

    @risharde said: Since op seemed to have recovered from the issue

    Working on it... there's some data corruption.

    @darkimmortal said: and use another tool to compare.

    Since the backup is encrypted, it'd have to download and decrypt all the data to compare it. At that point I may as well just restore the entire backup anyways. I realised that I actually only have ~1TB in the backup as most of the space used on the server is actually backups from other systems where I can just use borg check to check it.

    It looks like might be a way, borg create --files-cache disabled, not sure if that will give useful log output alone, if not followed withborg diff

Sign In or Register to comment.