zfs l2arc warmup

[Karl Denninger]

On 3/28/2014 4:45 PM, Dmitry Morozovsky wrote:
> On Fri, 28 Mar 2014, mikej wrote:
>
>>> [snip most again]
>>>
>>>> Around ZFSv14-ish, the ability to import a pool with a missing ZIL was
>>>> added.
>>>>
>>>> Remember the flow of data in ZFS:
>>>>    async write request --> TXG --> disk
>>>>    sync write request --> ZIL
>>>>                 \--> TXG --> disk
>>>>
>>>> All sync writes are written to the pool as part of a normal async TXG
>>>> after
>>>> its written sync to the ZIL.  And the ZIL is only ever read during pool
>>>> import.
>>> On the other side, doesn't it put the risk on sync-dependent, like database,
>>> systems?
>>>
>>> I'm thinking not about losing the transaction, but possibly putting your
>>> filesystem in the middle of (database PoV) transaction, hence render your DB
>>> inconsistent?
>>>
>>> Quick googling seems to be uncertain about it...
>> As I understand it..... (And I am always looking for an education)
>>
>> Any files system that honors fsync and provided the DB uses fsync should be
>> fine.
>>
>> Any data loss then will only be determined by what transaction (log)
>> capabilities the DB has.
> And?
>
> 1. The DB issues "sync WAL" request, which is translated to fsync-like FS
> requests, there are (IIUC) should ne directed to ZIL.
>
> 2. ZIL is failing in the middle of the request, or, even more bad, after
> reporting that ZIL transaction is done, but before translating ZIL to the
> underlying media
>
> 3. inconsistend DB?
>
> I'm in hope I'm wrong somewhere...
If the DB is EVER lied to on a fsync'd write (that is, it gets back 
completion when the write was not actually complete and on stable 
storage) you're asking for a corrupted database.

It doesn't much matter WHY the DB was lied to.  All modern database 
systems rely on the filesystem and operating system NOT lying to them 
about the fact that data written is on stable storage when a fsync'd 
write call returns.

-- 
-- Karl
karl at denninger.net


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2711 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20140328/875b55d1/attachment.bin>