Red Hat Bugzilla – Bug 1267242
Data Tiering feature of glusterfs requires update of sqlite version in RHEL6 to sqlite 3.7 or above to support WAL journaling mode
Last modified: 2016-01-25 16:11:40 EST
The WAL journaling mode does not work in sqlite 3.6.20-1 available at RHEL 6.
WAL Journaling mode is required for the Data Tiering feature of Red Hat Gluster Storage on RHEL 6
It is required to upgrade sqlite version on RHEL 6.7 to sqlite 3.7 and above, which supports WAL journaling mode
We measured the performance of our tiering feature with the current version of sqlite in RHEL 6.7 and it is very poor. We won't have a viable feature without sqlite 3.7 or higher. We need RHEL PM to weigh this request very closely as RHGS's tiering feature is not usable without a rebase of the sqlite version in RHEL. Our other alternative would be to carry a newer version of sqlite in our channels and we're trying to avoid doing that.
So we have a risk of issues on customers side (remember, we speak about data here, not only broken application) and performance improvement (how big?) on the other side. For me, bundling the sqlite with RHGS is the only viable choice here.
Is there no plan to rebase sqlite for the entire life-cycle of RHEL 6 i.e. till 2020? We waited for the RHEL 6.8 planning window to open to make our request.
Can someone please respond to my above question?
(In reply to Sayan Saha from comment #7)
> Is there no plan to rebase sqlite for the entire life-cycle of RHEL 6 i.e.
> till 2020? We waited for the RHEL 6.8 planning window to open to make our
Currently there is no such plan, which means the sqlite would stay in 3.6 till EOL of RHEL-6, yes. As already said, there are big risks that we would cause troubles by doing the rebase. From my PoV, the benefits are not worth such a big risk.
Created attachment 1100643 [details]
RHGS tiering performance delta when using a newer version of sqlite
Please find attached a graph that shows how big the performance difference is for the toering feature in RHGS if we use a newer version of sqlite.
Sayan, I see your point, the performance delta seems to be significant.
However, that doesn't change the statement that upgrading in RHEL 6 would be too risky. The upstream says :
"... Since 2004, there have been enhancements to SQLite such that newer database files are unreadable by older versions of the SQLite library..."
Such forward incompatibility ^^ is something we cannot afford in RHEL minor update, because it can easily lead to issues like in .
That said, I'd like to ask you to think seriously about some alternative solution, like having own release of sqlite for your product.